├── CONTRIBUTING.md
├── Deep Learning
├── Convolutionals Neural Network Guide.md
├── DL in General.md
├── Deep Learning For NLP.md
├── Transfer Learning.md
└── images
│ ├── Screenshot from 2018-09-30 08-21-45.png
│ ├── Screenshot from 2018-09-30 08-22-13.png
│ ├── Screenshot from 2018-09-30 08-43-37.png
│ ├── Screenshot from 2018-09-30 08-49-08.png
│ ├── Screenshot from 2018-09-30 09-28-06.png
│ ├── Screenshot from 2018-09-30 09-32-15.png
│ ├── Screenshot from 2018-09-30 09-36-03.png
│ ├── Screenshot from 2018-09-30 09-45-48.png
│ ├── Screenshot from 2018-09-30 10-00-46.png
│ ├── Screenshot from 2018-09-30 10-08-03.png
│ ├── Screenshot from 2018-09-30 10-08-35.png
│ └── Screenshot from 2018-09-30 10-13-24.png
├── LICENSE
├── ML in General.md
├── MLInterview Questions.md
├── Natural Language Processing.md
├── README.md
├── Reinforcement Learning.md
├── Statistics.md
├── Supervised Learning.md
├── Unsupervised Learning.md
└── images
├── 10_fold_cv.png
├── 1_7VyTVnlVj2Ooqa3MRSRFEQ.gif
├── 1_ET8kCcPpr893vNZFs8j4xg.gif
├── 1_KrcZK0xYgTa4qFrVr0fO2w.gif
├── 1_pdfGx9Hfo_01NIiL0A4iPA.png
├── 1_tEo4x2-2iOZcnhSF13rUTA.png
├── 20170513_gradient_descent_logistic_animation.gif
├── Adjusted_R2.png_large
├── An-illustration-of-neural-network-with-and-without-dropout.png
├── CBOW.png
├── Precisionrecall.svg.png
├── Screenshot from 2018-07-03 19-25-29.png
├── Screenshot from 2018-07-08 09-51-27.png
├── Screenshot from 2018-07-09 09-27-32.png
├── Screenshot from 2018-07-09 09-40-16.png
├── Screenshot from 2018-07-09 09-46-53.png
├── Screenshot from 2018-07-09 09-47-10.png
├── Screenshot from 2018-07-10 16-29-57.png
├── Screenshot from 2018-08-12 09-28-06.png
├── Screenshot from 2018-08-21 20-22-44.png
├── Screenshot from 2018-08-21 20-24-08.png
├── Screenshot from 2018-08-21 20-53-48.png
├── Screenshot from 2018-08-21 20-56-24.png
├── Screenshot from 2018-09-21 09-20-55.png
├── Screenshot from 2018-10-04 14-14-08.png
├── Screenshot from 2018-10-04 14-14-25.png
├── Screenshot from 2018-10-04 14-14-52.png
├── Screenshot from 2018-10-04 14-16-00.png
├── Screenshot from 2018-10-04 14-17-53.png
├── Screenshot from 2019-07-22 19-26-50.png
├── Screenshot from 2019-07-22 19-31-21.png
├── WordEmbeddings.png
├── a.png
├── b.png
├── bag.jpg
├── c.jpeg
├── d.png
├── decision-forests-and-discriminant-analysis-77-638.jpg
├── download (1).jpeg
├── download.jpeg
├── download.png
├── e.png
├── edge network.png
├── f.png
├── g.jpeg
├── g.png
├── h.png
├── healthreflect.png
├── i.jpeg
├── i.png
├── image5.png
├── image6-e1536165830511.png
├── inbox_2181887_2db081ae1af423307f7ae558f76a2063_silhoutte.png
├── inbox_2181887_373d4f8bda1be2b867c2f404dd335e85_unnamed-chunk-16-1.png
├── inbox_2181887_42dfac7118608f18d67365a3c35a7aa6_Elbow method.png
├── indigene.png
├── innoplexus.png
├── j.jpeg
├── kmeansclusteringalgo.png
├── maxresdefault.jpg
├── me-machine-learning-mathematics-so-true-41965149.png
├── partitions-underfitting-vs-overfitting-regression-via-polynomial-degree.png
├── roc.gif
├── seashell.png
├── svmhyperplane.png
└── workflow.png
/CONTRIBUTING.md:
--------------------------------------------------------------------------------
1 | ### Contributing
2 |
3 | If you would like to add a new AI startups, you can do so with a pull request (PR).
4 | The startup should include the name of the company, in which year it was found, and a link to the careers page of the company with a short note about company.
5 |
6 | `Name` column (see below) to the table if it does not exist.
7 | `Products & Careers Page` column give an overview about startup and link to the careers page of the company.
8 |
9 | | Name | Product & Careers Page | Year Founded |
10 | | ------------- | :-----:| --- |
11 | |Name of the company | Short Overview [Link to the Careers Page] | 2013 |
12 |
13 | 1. Fork the repository.
14 | 2. Make the changes required.
15 | 3. Submit your change as a pull request.
16 |
--------------------------------------------------------------------------------
/Deep Learning/Convolutionals Neural Network Guide.md:
--------------------------------------------------------------------------------
1 | # Convolutional Neural Network
2 | The AlexNet, was first able to beat more traditional geometrical approaches on the most popular object recognition contest - the ILSVRC. Ever since, DCNNs have been achieving state-of-the-art results on object recognition tasks.
3 |
4 | # Why CNN
5 |
6 | If you would use Fully connected ANN then you would have to train a lot of parameters and network has to optimize all of these parameters,
7 | the training process could then become very time and storage intensive. In order to solve this computational problem, a different
8 | kind of network architecture is used, called **Convolutional Neural Network (CNN)**
9 |
10 | CNNs introduce two new types of hidden layers in which each neuron is only connected to a small subset of the neurons in the previous layer, to prevent the
11 | aforementioned problem.
12 |
13 | 1. **Convolutional Layer**: What separates a convolutional layer from a fully-connected one is that each neuron is only connected to a small, local subset of the
14 | neurons in the previous layer, which is a square sized region across the height and width dimensions. The size of this square
15 | is a hyperparameter named **Receptive Field**.
16 |
17 | 
18 |
19 | Neurons of the convolution operator can recognize certain local patterns of the previous layer’s output.In order to now reconize multiple different features within
20 | one layer, it is required to have several Filters. How many convolutions are being conducted is defined by another hyperparameter, called Stride, which determines
21 | how big the gap between two scanned regions is.
22 |
23 | Why Padding ?? (Having Certain Width & Height + Cover the edges of images much better )
24 |
25 | 2. **Pooling Layer**: The third kind of layer, which has the purpose of decreasing the complexity of CNNs, is the Pooling
26 | Layer. Similarly to the convolutional layer, neurons in the pooling layer are connected to a square sized region across
27 | the width and height dimensions of the previous layer. The main difference between convolution and pooling is that a
28 | pooling layer is not Parametrized.
29 |
30 | * Max Pooling
31 | * Average Pooling
32 |
33 | 
34 |
35 | ## APPLICATIONS OF DCNN FOR OBJECT RECOGNITION TASKS
36 |
37 | Three object recognition tasks - classification, localization and detection - and how each of them can be tackled with DCNNs.
38 |
39 | ### Classification
40 |
41 | The task of Image Classification describes the challenge of categorizing a given image into one of several classes. A
42 | possible application of this could be the recognition of hand-written digits, where the input image is classified as one of the
43 | ten classes.
44 |
45 | 
46 |
47 | ### Localization
48 |
49 | For Localization, the information about which category an image belongs to is already available and the task is to instead
50 | figure out where exactly the object is located in the image. This location is typically specified by a two-dimensional bounding
51 | box, which consists of four values that describe the location of two opposite couples of corners. Finding these four values is
52 | the main challenge of localization and is commonly referred to as **Bounding Box Regression**.
53 |
54 | 
55 |
56 | To perform a localization task, we can use a similar architecture as the one we defined for classification. The only
57 | thing that has to be modified is the final output layer, which can simply be replaced by another output layer that performs
58 | the bounding box regression instead.
59 |
60 | ### Detection
61 |
62 | In contrast to multi-class localization, the number of objects in a given image is not known prior to the execution, when performing Object Detection. In order to use DCNNs for such
63 | object detection tasks, the architecture needs to be extended to handle the flexible amount of detections.
64 |
65 | * **R-CNN**: **Region-based Convolutional Neural Networks (R-CNNs)** are using Region Proposal Networks to only extract potentially interesting regions. These regions are called
66 | Regions of Interest (ROIs). They are obtained by running a quick segmentation to spot blob-like structures.
67 |
68 |
69 | * **R-FCN**: These networks are structured very similarly to the faster R-CNN architecture, but instead of using fully-
70 | connected modules to predict classes and bounding boxes for each ROI, R-FCNs use Position-Sensitive Convolutional Modules.
71 |
72 | * **YOLO** (You Only Look Once): In contrast with region proposal based techniques, Single-Shot detection architectures do not predict any
73 | regions of interests, but instead, a fixed amount of detections on the image directly, which are then filtered to contain only
74 | the actual detections. These networks do therefore have much faster execution times than region-based architectures but are
75 | found to also have a lower detection accuracy.
76 |
77 | Note : Suitable for real time applications.
78 | 
79 |
80 | * **Single Shot MultiBox Detectors (SSD)** : In order to handle the problems of YOLO that arise due to the fixed amount of predictions and cell sizes,
81 | Single Shot MultiBox Detectors (SSD) have been developed, which predict detections of different scales and also make
82 | predictions for multiple different aspect ratios. Therefore, SSD detectors can make finer predictions, leading to significantly
83 | better results.
84 |
85 |
86 | ## DCNN ARCHITECTURES
87 |
88 | ImageNet Large Scale Visual Recognition Competition (ILSVRC) - Models in chronological order.
89 |
90 | * **LeNet-5 (1998)** [3 CL + 2 Pooling + 1 FC]
91 |
92 | Most of the DCNNs that are being used for object recognition today are based on this basic architecture. As can be seen, the network architecture is relatively simple,
93 | as it only consists of an input layer of size 32 × 32, an output layer of size 10, as well as three 5 × 5 convolutional, two
94 | 2×2 pooling and one fully-connected layer in between, making it a total of six hidden layers.
95 | 
96 |
97 | * **AlexNet (2012)** [5 CL + 3 Pooling + 2 FC]
98 |
99 | It was the first DCNN that managed to beat more traditional object recognition approaches in the ILSVRC.
100 | 
101 |
102 | * ZFNet and OverFeat (2013) [Same as AlexNet but with a slight modification]
103 |
104 | * **VGGNet and GoogLeNet (2014)**
105 |
106 | VGGNet, scored second place in the ILSVRC and influenced the deep learning scene in an important way, as they showed that using a deeper architecture
107 | does generally lead to better results, which was not obvious at that time.
108 | 
109 |
110 | Google won the ILSVRC with a different very deep network that had 22 parametrized layers - even more than the VGGNet - and was named GoogLeNet. This network was an improvement of the AlexNet that was not only much deeper but also
111 | reduced the number of parameters. The latter was achieved by replacing the first fully-connected layer, which is typically
112 | accountable for the highest number of parameters, by another convolutional layer. In addition to that, they also implemented
113 | the so-called Inception Modules, which enable a network to recognize patterns of different sizes within the same layer.
114 | 
115 |
116 | * **ResNet** (2015)
117 |
118 | Their network, called Deep Residual Network, or ResNet for short, is able to perform much better with very deep architectures. In
119 | ResNets, convolutional layers are divided into Residual Blocks and for each block a Residual Connections is added, which is
120 | bypassing the corresponding block. By adding these residual connections, the result of a training
121 | step can be backpropagated to the earlier layers directly, without any interference from subsequent layers. Therefore, resid-
122 | ual connections enable the training of even deeper networks.
123 |
124 | 
125 |
126 |
127 | * Inception-v4, Inception-ResNet-v1 and ResNeXt (2016)
128 |
129 | * Densenet, DPN and MobileNets (2017)
130 |
131 |
132 |
133 | Refrences:
134 |
135 | * [A Non-Technical Survey on Deep Convolutional Neural Network Architectures](https://arxiv.org/pdf/1803.02129.pdf):star::star::star::star:
136 | * [CS231n: Convolutional Neural Networks for Visual Recognition](http://cs231n.stanford.edu/)
137 |
--------------------------------------------------------------------------------
/Deep Learning/DL in General.md:
--------------------------------------------------------------------------------
1 | # Deep Learning
2 |
3 | ## Activation Function
4 |
5 | * Sigmoid (not used because they suffer from vanishing and exploding gradient problem, except in LSTM/RNN)
6 | * Tanh
7 | * ReLu (Dead Neurons Problem)
8 | * Leaky ReLu (Solve the Dead Neurons Problem)
9 | * ELU (Exponential Linear Unit)
10 |
11 | ## Dropout
12 |
13 | A form of regularization useful in training neural networks. Dropout regularization works by removing a random selection of a fixed number of the units in a network layer for a single gradient step. The more units dropped out, the stronger the regularization. This is analogous to training the network to emulate an exponentially large ensemble of smaller networks.
14 |
--------------------------------------------------------------------------------
/Deep Learning/Deep Learning For NLP.md:
--------------------------------------------------------------------------------
1 | # Deep Learning for Natural Language Processing
2 |
3 |
--------------------------------------------------------------------------------
/Deep Learning/Transfer Learning.md:
--------------------------------------------------------------------------------
1 | # Transfer Learning
2 |
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 08-21-45.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 08-21-45.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 08-22-13.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 08-22-13.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 08-43-37.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 08-43-37.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 08-49-08.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 08-49-08.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 09-28-06.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 09-28-06.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 09-32-15.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 09-32-15.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 09-36-03.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 09-36-03.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 09-45-48.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 09-45-48.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 10-00-46.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 10-00-46.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 10-08-03.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 10-08-03.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 10-08-35.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 10-08-35.png
--------------------------------------------------------------------------------
/Deep Learning/images/Screenshot from 2018-09-30 10-13-24.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/Deep Learning/images/Screenshot from 2018-09-30 10-13-24.png
--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
1 | MIT License
2 |
3 | Copyright (c) 2018 SHYAM SUNDER KUMAR
4 |
5 | Permission is hereby granted, free of charge, to any person obtaining a copy
6 | of this software and associated documentation files (the "Software"), to deal
7 | in the Software without restriction, including without limitation the rights
8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9 | copies of the Software, and to permit persons to whom the Software is
10 | furnished to do so, subject to the following conditions:
11 |
12 | The above copyright notice and this permission notice shall be included in all
13 | copies or substantial portions of the Software.
14 |
15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 | SOFTWARE.
22 |
--------------------------------------------------------------------------------
/ML in General.md:
--------------------------------------------------------------------------------
1 | # Table of Contents
2 | * [Types of Artificial Intelligence Learning Models](#aimodels)
3 | * [Knowledge-Based Classification](#kbc)
4 | * [Feedback-Based Classification](#fbc)
5 | * [Data Mining Vs Machine Learning](#mlvsdm)
6 | * [Workflow of Machine Learning Project](#workflow)
7 | * [Parametric vs Nonparametric](#tric)
8 | * [Discriminative vs Generative Learning Algorithm](#dam)
9 | * [Cross Validation](#cv)
10 | * [Overfitting](#of)
11 | * [Regularization](#reg)
12 | * [Loss Functions for Regression and Classification](#lossfun)
13 | * [Handle missing or Corrupted data](#missing)
14 | * [Imbalanced Datasets](#imb)
15 | * [Outliers](#out)
16 |
17 | ## Types of Artificial Intelligence Learning Models
18 |
19 | ### AI Learning Models: Knowledge-Based Classification
20 | - **Inductive Learning**: This type of AI learning model is based on inferring a general rule from datasets of input-output pairs.
21 | - **Deductive Learning**: This type of AI learning technique starts with a series of rules and infers new rules that are more efficient in the context of a specific AI algorithm.
22 |
23 | ### AI Learning Models: Feedback-Based Classification
24 |
25 | Based on the feedback characteristics, AI learning models can be classified as supervised, unsupervised, semi-supervised or reinforced.
26 |
27 | — **Unsupervised Learning**: Unsupervised models focus on learning a pattern in the input data without any external feedback. Clustering is a classic example of unsupervised learning models.
28 |
29 | — **Supervised Learning**: Supervised learning models use external feedback to learning functions that map inputs to output observations. In those models the external environment acts as a “teacher” of the AI algorithms.
30 |
31 | — **Semi-supervised Learning**: Semi-supervised learning is a class of supervised learning tasks and techniques that also make use of unlabeled data for training – typically a small amount of labeled data with a large amount of unlabeled data.
32 | The goal of a semi-supervised model is to classify some of the unlabeled data using the labeled information set.
33 |
34 | — **Reinforcement Learning**: Reinforcement learning models use opposite dynamics such as rewards and punishment to “reinforce” different types of knowledge. This type of learning technique is becoming really popular in modern AI solutions.
35 |
36 | ## Data Mining Vs Machine Learning
37 |
38 | **Machine learning** focuses on prediction, based on known properties learned from the training data.
39 | **Data mining** focuses on the discovery of (previously) unknown properties in the data. This is the analysis step of Knowledge Discovery in Databases.
40 |
41 | **Note** : If you have a better explanation a pull request would be helpful.
42 |
43 | ## Workflow of Data Science Project
44 |
45 | Given a data science / machine learning project, what steps should we follow? Here's how we should tackle it:
46 |
47 | * **Specify business objective.** Are we trying to win more customers, achieve higher satisfaction, or gain more revenues?
48 | * **Define problem.** What is the specific gap in your ideal world and the real one that requires machine learning to fill? Ask questions that can be addressed using your data and predictive modeling (ML algorithms).
49 | * **Create a common sense baseline.** But before you resort to ML, set up a baseline to solve the problem as if you know zero data science. You may be amazed at how effective this baseline is. It can be as simple as recommending the top N popular items or other rule-based logic. This baseline can also server as a good benchmark for ML algorithms.
50 | * **Review ML literatures.** To avoid reinventing the wheel and get inspired on what techniques / algorithms are good at addressing the questions using our data.
51 | * **Set up a single-number metric.** What it means to be successful - high accuracy, lower error, or bigger AUC - and how do you measure it? The metric has to align with high-level goals, most often the success of your business. Set up a single-number against which all models are measured.
52 | * **Do exploratory data analysis (EDA).** Play with the data to get a general idea of data type, distribution, variable correlation, facets etc. This step would involve a lot of plotting.
53 | * **Partition data.** Validation set should be large enough to detect differences between the models you are training; test set should be large enough to indicate the overall performance of the final model; training set, needless to say, the larger the merrier.
54 | * **Preprocess.** This would include data integration, cleaning, transformation, reduction, discretization and more.
55 | * **Engineer features.** Coming up with features is difficult, time-consuming, requires expert knowledge. Applied machine learning is basically feature engineering. This step usually involves feature selection and creation, using domain knowledge. Can be minimal for deep learning projects.
56 | * **Develop models.** Choose which algorithm to use, what hyperparameters to tune, which architecture to use etc.
57 | * **Ensemble.** Ensemble can usually boost performance, depending on the correlations of the models/features. So it’s always a good idea to try out. But be open-minded about making tradeoff - some ensemble are too complex/slow to put into production.
58 | * **Deploy model.** Deploy models into production for inference.
59 | * **Monitor model.** Monitor model performance, and collect feedbacks.
60 | * **Iterate.** Iterate the previous steps. Data science tends to be an iterative process, with new and improved models being developed over time.
61 |
62 | 
63 |
64 | ## Parametric vs Nonparametric ?
65 | A learning model that summarizes data with a set of parameters of fixed size (independent of the number of training examples) is called a **parametric model**.
66 | A learning model where the number of parameters is not determined prior to training. On the contrary, nonparametric models (can) become more and more complex with an increasing amount of data.
67 |
68 | ## Discriminative vs Generative Learning Algorithm ?
69 | **Discriminative algorithms model p(y|x; w)**, that is, given the dataset and learned parameter, what is the probability of y belonging to a specific class. A discriminative algorithm doesn't care about how the data was generated, it simply categorizes a given example
70 | Ex: Linear Regression, Logistic Regression, Support Vector Machines etc.
71 |
72 | **Generative algorithms model p(x|y)**, that is, the distribution of features given that it belongs to a certain class. A generative algorithm models how the data was generated.
73 | Ex: Naive Bayes, Hidden Markov Models etc.
74 |
75 | 
76 |
77 | Given a training set, an algorithm like logistic regression or the perceptron algorithm (basically) tries to find a straight line—that is, a decision boundary—that separates the elephants and dogs. Then, to classify a new animal as either an elephant or a dog, it checks on which side of the decision boundary it falls, and makes its prediction accordingly.
78 |
79 | First, looking at elephants, we can build a model of what elephants look like. Then, looking at dogs, we can build a separate model of what dogs look like. Finally, to classify a new animal, we can match the new animal against the elephant model, and match it against the dog model, to see whether the new animal looks more like the elephants or more like the dogs we had seen in the training set.
80 |
81 | ## What is cross validation ?
82 |
83 | Cross Validation is a technique to evaluate predictive models by partitioning the original sample into a training set to train the model, and a validation set to evaluate it. For ex: K fold CV divides the data into k folds, train on each k-1 folds and evaluate it on remaining 1 fold. The result of k models can be averaged to get a overall model performance.
84 |
85 | 
86 |
87 | **Time - Series Cross Validation** : Experimenters cannot cut out a piece in the middle, and train on data before and after this portion. Instead, they need to train on a set of data that is older than the test data.
88 | 
89 |
90 | With this in mind, there are two major approaches, outlined in Figure, above: the **sliding window** approach and the **expanding window** approach. In the sliding window approach, one uses a fixed size window, shown here in black, for training. Subsequently, the method is tested against the data shown in orange.
91 |
92 | ## What is overfitting?
93 |
94 | Overfitting or High Variance is a modeling error which is caused by a hypothesis function that fits the training data too close but does not generalise well to predict new data.
95 | 
96 |
97 | ## What is regularization?
98 |
99 | Regulariztion is a technique to prevent overfitting by penalizing the coefficients of the cost function.
100 |
101 | ### Ridge Regression
102 | It performs ‘L2 regularization’, i.e. adds penalty equivalent to square of the magnitude of coefficients.
103 | `L 2 regularizer` is also called a gaussian prior or weight decay .
104 | Thus, it optimises the following:
105 |
106 | Objective = RSS + α * (sum of square of coefficients)
107 |
108 | ### Lasso Regression
109 | LASSO stands for Least Absolute Shrinkage and Selection Operator.Lasso regression performs L1 regularization, i.e. it adds a factor of sum of absolute value of coefficients in the optimisation objective.
110 |
111 | Objective = RSS + α * (sum of absolute value of coefficients)
112 |
113 | ### Elastic nets
114 | A technique known as Elastic Nets, which is a combination of Lasso
115 | and Ridge regression is used to tackle the limitations of both Ridge and
116 | Lasso Regression.
117 |
118 |
119 | ## Loss Functions for Regression and Classification?
120 |
121 | * **Regression Loss Function**
122 | * Square or l2 loss (not robust)
123 | * Absolute or Laplace loss (not differentiable)
124 | * Huber Loss (robust and differentiable)
125 | * **Classification Loss Function**
126 | * SVM/Hinge loss
127 | * log loss
128 |
129 | 
130 |
131 | ## How do you handle missing or corrupted data in a dataset?
132 | Before jumping to the methods of data imputation, we have to understand the reason why data goes missing.
133 |
134 | - **Missing Completely at Random (MCAR)**: The fact that a certain value is missing has nothing to do with its hypothetical value and with the values of other variables.
135 |
136 | - **Missing at Random (MAR) - a weaker assumption than MCAR**: Missing at random means that the propensity for a data point to be missing is not related to the missing data, but it is related to some of the observed data.
137 |
138 | - **Missing not at Random (MNAR)**: Two possible reasons are that the missing value depends on the hypothetical value (e.g. People with high salaries generally do not want to reveal their incomes in surveys) or missing value is dependent on some other variable’s value (e.g. Let’s assume that females generally don’t want to reveal their ages! Here the missing value in age variable is impacted by gender variable).
139 |
140 | Methods :
141 |
142 | * **Listwise Deletion** : In the listwise deletion method, all rows that have one or more column values missing are deleted.
143 |
144 | * **Mean, Median and Mode Imputation**: In the mean/median/mode imputation method, all missing values in a particular column are substituted with the mean/median/mode, which is calculated using all the values available in that column.
145 |
146 | * Multiple Imputation
147 | * Last Observation Carried Forward (LOCF)**
148 | * KNN (K Nearest Neighbors)
149 |
150 | ## How would you handle an imbalanced dataset?
151 | * Using a better metrics like AUROC, Precision, Recall etc.
152 | * Cost-sensitive Learning
153 | * Over sampling of the minority class or Under sampling of the majority class.
154 | * SMOTE (Synthetic Minority Over-sampling Technique.)
155 | * Anomaly Detection
156 |
157 | 
158 |
159 | ## how do you detect outliers?
160 |
161 | Outliers are extreme values that deviate from other observations on data , they may indicate a variability in a measurement, experimental errors or a novelty.
162 |
163 | ### How to find outliers
164 | * Visualize the Data
165 | * **Histogram**: A histogram is the best way to check univariate data — data containing a single variable — for outliers
166 | * **Scatter Plot**: A scatter plot is useful to find outliers in bivariate data (data with two variables). You can easily spot the outliers because they will be far away from the majority of points on the scatter plot.
167 | * **Box Plots**
168 | * incomplete
169 |
170 |
--------------------------------------------------------------------------------
/MLInterview Questions.md:
--------------------------------------------------------------------------------
1 | # Machine Learning Interview Questions
2 |
3 | ## Cross Entropy or Log Loss
4 |
5 | Cross-entropy is commonly used to quantify the difference between two probability distributions.
6 | Cross-entropy loss measures how close is the predicted distribution to the true distribution.
7 |
8 | 
9 |
10 | Why the Negative Sign?
11 | Log Loss uses negative log to provide an easy metric for comparison. It takes this approach because the positive log of numbers < 1 returns negative values, which is confusing to work with when comparing the performance of two models.
12 |
13 | ## Explain Bias-Variance Tradeoff
14 |
15 | **Bias** is the simplifying assumptions made by the model to make the target function easier to approximate.
16 |
17 | **Variance** is the amount that the estimate of the target function will change given different training data.
18 |
19 | The **bias–variance trade-off** is the conflict in trying to simultaneously minimize these two sources of error that prevent supervised learning algorithms from generalizing beyond their training set.
20 |
21 | 
22 |
23 | ## References
24 |
25 | [Machine Learning Mastery- Bias Variance Tradeoff](https://machinelearningmastery.com/gentle-introduction-to-the-bias-variance-trade-off-in-machine-learning/)
26 |
27 |
--------------------------------------------------------------------------------
/Natural Language Processing.md:
--------------------------------------------------------------------------------
1 | # Natural Language Processing
2 |
3 | ## How to clean your text data?
4 |
5 | - Remove all irrelevant characters such as any non alphanumeric characters
6 | - Tokenize your text by separating it into individual words
7 | - Remove words that are not relevant, such as “@” twitter mentions or urls
8 | - Convert all characters to lowercase((**Case folding**), in order to treat words such as “hello”, “Hello”, and “HELLO” the same. (Note)
9 | - Consider combining misspelled or alternately spelled words to a single representation (e.g. “cool”/”kewl”/”cooool”)
10 | - Consider lemmatization (reduce words such as “am”, “are”, and “is” to a common form such as “be”)
11 | - Consider removing stopwords (such as a, an, the, be)etc.
12 |
13 | Note : For tasks like speech recognition and information retrieval, everything is mapped to lower case. For sentiment anal-
14 | ysis and other text classification tasks, information extraction, and machine translation, by contrast, case is quite helpful and case folding is generally not done (losing the difference, for example, between US the country and us the pronoun can out-
15 | weigh the advantage in generality that case folding provides)
16 |
17 | ## What is Tokenization?
18 |
19 | **Tokenization** is the process of converting a sequence of characters into a sequence of tokens.
20 | Ex :RegexpTokenizer & Word Tokenize (scikit-learn)
21 | 
22 |
23 | ## What is stop words?
24 |
25 | **Stop words** are words that are particularly common in a text corpus and thus considered as rather un-informative.
26 | 
27 | ## What is Stemming and lemmatization?
28 |
29 | The goal of both stemming and lemmatization is to reduce inflectional forms and sometimes derivationally related forms of a word to a common base form.
30 |
31 | **Stemming** usually refers to a crude heuristic process that chops off the ends of words in the hope of achieving this goal correctly most of the time, and often includes the removal of derivational affixes. Different types of stemmers in NLTK are PorterStemmer, LancasterStemmer, SnowballStemmer.
32 | 
33 |
34 | **Lemmatization** usually refers to doing things properly with the use of a vocabulary and morphological analysis of words, normally aiming to remove inflectional endings only and to return the base or dictionary form of a word, which is known as the lemma.
35 | 
36 |
37 | Note : It uses a knowledgebase called WordNet. Because of knowledge, lemmatization can even convert words which are different and cant be solved by stemmers, for example converting “came” to “come”.
38 |
39 | ## Parts-of-Speech (POS) tagging
40 |
41 | Part-of-speech tagging (POS tagging) is the task of tagging a word in a text with its part of speech. A part of speech is a category of words with similar grammatical properties. Common English parts of speech are noun, verb, adjective, adverb, pronoun, preposition, conjunction, etc.
42 |
43 | ## Named Entity Recognition (NER)
44 |
45 | In the Named Entity Recognition (NER) task, systems are required to recognize the Named Entities occurring in the text. More specifically, the task is to find Person (PER), Organization (ORG), Location
46 | (LOC) and Geo-Political Entities (GPE). For instance, in the statement ”Shyam lives in India”, NER system extracts Shyam which refers to name of the person and India which refers to name of the country.
47 |
48 | ## Corefrence resolution(CR)
49 |
50 | Coreference Resolution is the task which determines which noun phrases (including pronouns,
51 | proper names and common names) refer to the same entities in documents . For instance, in the sentence, ”I have seen the annual report. It shows that we have gained 15% profit in this financial year”. Here, ”I” refers to name of the person, ”It” refers to annual report and ”we” refers to the name of the company in which that person works.(Kong et al., 2010)
52 |
53 | ## What is Bag of Words?
54 | **Bag of words (BoW)** builds a vocabulary of all the unique words in our dataset, and associate a unique index to each word in the vocabulary.It is called a "bag" of words, because it is a representation that completely ignores the order of words.
55 | 
56 |
57 | ## What is tf - idf?
58 | **TF-IDF** reveals what words are the most discriminating between different bodies of text. It is dependent on term frequency, how often a word appears, and Inverse document frequency, whether it is unique or common among all documents. It is particularly, helpful if you are trying to see the difference between words that occur a lot in one document, but fail to appear in others allowing you interpret something special about that document.
59 |
60 | **Example**:
61 |
62 | Consider a document containing 100 words wherein the word cat appears 3 times. The term frequency (i.e., tf) for cat is then (3 / 100) = 0.03. Now, assume we have 10 million documents and the word cat appears in one thousand of these. Then, the inverse document frequency (i.e., idf) is calculated as log(10,000,000 / 1,000) = 4. Thus, the Tf-idf weight is the product of these quantities: 0.03 * 4 = 0.12.
63 |
64 | ## N gram
65 |
66 | * n-gram is a contiguous sequence of n items from a given sample of text or speech
67 | * An n-gram of size 1 is referred to as a "unigram"; size 2 is a "bigram" size 3 is a "trigram". Larger sizes are sometimes referred to by the value of n in modern language, e.g., "four-gram", "five-gram", and so on.
68 | 
69 | * ngram model models sequence, i.e., predicts next word (n) given previous words (1, 2, 3, ..., n-1)
70 | * multiple gram (bigram and above) captures **context**
71 | * to choose n in n-gram requires experiments and making tradeoff between stability of the estimate against its appropriateness. Rule of thumb: trigram is a common choice with large training corpora (millions of words), whereas a bigram is often used with smaller ones.
72 | * n-gram can be used as features for machine learning and downstream NLP tasks
73 |
74 | ## What is word2vec ?
75 |
76 | It is a shallow two-layer neural networks that are trained to construct linguistic context of words.
77 | It Takes as input a large corpus, and produce a vector space, typically of several hundred dimension, and each word in the corpus is assigned a vector in the space.
78 | The key idea is context: words that occur often in the same context should have same/opposite meanings.
79 | 
80 |
81 | Two Algorithms:
82 | * **Skip-Grams** : The skip-gram model does the exact opposite of the CBOW model, by predicting the surrounding context words
83 | given the central target word.
84 | * **Continuous Bag of Words (CBOW)**: CBOW computes the conditional probability of a target word given the context words surrounding it across a window of size k.
85 |
86 | **Limitations**:
87 |
88 | * When we want to obtain vector representations for phrases such as “hot potato” or “Boston Globe”. We can’t just simply combine the individual word vector representations since these phrases don’t represent the combination of meaning of the individual words. And it gets even more complicated when longer phrases and sentences are considered.
89 | * use of smaller window sizes produce similar embeddings for contrasting words such as “good” and “bad”, which is not desirable especially for tasks where this differentiation is important such as sentiment analysis.At times these embeddings cluster semantically-similar words which have opposing sentiment polarities.
90 |
91 |
92 |
93 | 
94 |
95 | ## NLP Metrics
96 |
97 | **Perplexity** : The perplexity (sometimes called PP for short)
98 | of a language model on a test set is the inverse probability of the test set, normalized
99 | by the number of words. (*Smaller is better*)
100 |
101 | Note : Perplexity does not guarantee an (extrinsic) im-
102 | provement in the performance of a language processing task like speech recognition
103 | or machine translation.
104 |
105 |
106 |
--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
1 | # Machine Learning & Data Science Interview [](https://github.com/theainerd/MLInterview/blob/master/LICENSE)
2 |
3 | [](https://sourcerer.io/fame/theainerd/theainerd/MLInterview/links/0)[](https://sourcerer.io/fame/theainerd/theainerd/MLInterview/links/1)[](https://sourcerer.io/fame/theainerd/theainerd/MLInterview/links/2)[](https://sourcerer.io/fame/theainerd/theainerd/MLInterview/links/3)[](https://sourcerer.io/fame/theainerd/theainerd/MLInterview/links/4)[](https://sourcerer.io/fame/theainerd/theainerd/MLInterview/links/5)[](https://sourcerer.io/fame/theainerd/theainerd/MLInterview/links/6)[](https://sourcerer.io/fame/theainerd/theainerd/MLInterview/links/7)
4 |
5 | > When you cease to dream, you cease to live.
6 |
7 | [Data Science Interview Prep](https://in.udacity.com/course/data-science-interview-prep--ud944)
8 |
9 | ## Statistics & Probability
10 | * [Intro-to-Descriptive-Statistics](https://towardsdatascience.com/intro-to-descriptive-statistics-252e9c464ac9)
11 | * [How To Ace Data Science Interviews: Statistics](https://towardsdatascience.com/how-to-ace-data-science-interviews-statistics-f3d363ad47b)
12 | ## Machine Learning
13 |
14 | ### Data Cleaning/Exploratory Analysis
15 | * [Exploratory Data Analysis](https://www.youtube.com/watch?v=zHcQPKP6NpM&t=247s)
16 | * [How to Handle Missing Data](https://towardsdatascience.com/how-to-handle-missing-data-8646b18db0d4)
17 | * [Unboxing Outliers In Machine Learning](https://medium.com/datadriveninvestor/unboxing-outliers-in-machine-learning-d43fe40d88a6)
18 |
19 | * [Regularization] [Regularization. Clarified.](https://medium.com/data-science-group-iitr/regularization-a-smooth-trick-to-increase-robustness-of-regression-models-a5e8a91737ff)
20 |
21 | ### Metrics
22 |
23 | * [Choosing the Right Metric for Evaluating ML Models — Part 1](https://towardsdatascience.com/choosing-the-right-metric-for-machine-learning-models-part-1-a99d7d7414e4)
24 | * [Choosing the Right Metric for Evaluating Machine Learning Models — Part 2](https://towardsdatascience.com/choosing-the-right-metric-for-evaluating-machine-learning-models-part-2-86d5649a5428)
25 |
26 | ## Deep Learning
27 | * [Convolutional Neural Networks (CNNs / ConvNets)](http://cs231n.github.io/convolutional-networks/)
28 | ## Natural Language Processing
29 | * [How to solve 90% of NLP problems: a step-by-step guide](https://blog.insightdatascience.com/how-to-solve-90-of-nlp-problems-a-step-by-step-guide-fda605278e4e)
30 |
31 | ## Computer Vision
32 |
33 | ## Programming
34 | * [How To Ace Data Science Interviews: R/Python](https://towardsdatascience.com/how-to-ace-data-science-interviews-r-python-3a49982000de)
35 | ## Others
36 |
37 | * [SQL] [How To Ace Data Science Interviews: SQL](https://towardsdatascience.com/how-to-ace-data-science-interviews-sql-b71de212e433)
38 | * [Learn SQL] [SQL Scavenger Hunt Handbook](https://www.kaggle.com/rtatman/sql-scavenger-hunt-handbook)
39 | * [Blogs] [May Edition: Getting A Job In Data Science](https://towardsdatascience.com/may-edition-getting-a-job-in-data-science-125996b1734c)
40 |
41 | ## These tips will keep you way ahead in the race
42 | * [Tips] [Ten Job Search Hacks Everybody Needs To Know](https://humanworkplace.com/blog/ten-job-search-hacks-everybody-needs-to-know)
43 | * [Cover Letter Advice] [Your Cover Letter Should Be About Me, Not You](https://medium.com/@janetktaylor/your-cover-letter-should-be-about-me-not-you-3f9c0c21773f)
44 | * [Refresh Your Resume by Google](https://in.udacity.com/course/refresh-your-resume--ud243)
45 |
46 | ## Machine Learning/ Analytics Startups in India (Careers Page)
47 |
48 | | Name | Product & Careers Page | Year Founded|
49 | | :---: | :---: | :---: |
50 | |  **Arya.ai** | Simplifies buildout and manages deployment of Deep Learning based applications [Arya.ai Careers](https://angel.co/arya-ai/jobs) | 2013 |
51 | |  **Haptik.ai** | World's largest chatbot platforms, building applications for consumers, publishers, and enterprises. [Haptik.ai Careers](https://haptik.ai/careers/) | 2013 |
52 | |  **Edge Networks** | NextGen Technology Solutions for Workforce Transformation [Edge Networks Careers](https://edgenetworks.in/job-opportunities/) | 2012 |
53 | |  **Netradyne** | Creating Solutions for real world problems using Computer Vision and Deep Learning Technologies [Netradyne Careers](https://netradyne.com/careers/) | 2015|
54 | |  **niki.ai** | Bot that will guide you along with recommendations to reach the right service & make transactions [niki.ai Careers](https://angel.co/niki/jobs) | 2015 |
55 | |  **Active.ai** | Delivers conversational banking services to help banks and financial institutions redefine their future digital strategy. [Active.ai Careers](https://active.ai/careers/) | 2016|
56 | |  **AIndra Systems** | AIndra Systems is a technology ai startup that builds products in the artificial intelligence domain [AINDRA Careers](http://www.aindra.in/) | 2012 |
57 | |  **Locus.sh** | Locus.sh uses data driven algorithms to optimize logistics for businesses. [Locus.sh Careers](https://locus.sh/careers/) | 2015 |
58 | |  **Mad Street Den** |Computer Vision for fashion under the brand [Mad Street Den Careers](https://angel.co/mad-street-den/jobs) | 2013 |
59 | |  **SocialCops** | Apps to collect data, visualization platforms [Social Cops Careers](https://socialcops.com/careers/culture/) | 2013 |
60 | |  **Morph.ai** | Morph uses a chatbot to create personalized conversions for marketing. [Morph.ai Careers](https://blog.morph.ai/careers-morph-ai-67f3090efc0e/) | 2016|
61 | |  **SigTuple** | SigTuple analyze visual medical data through artificial intelligence to aid diagnosis using intelligent screening solutions. [SigTuple Careers](https://sigtuple.bamboohr.com/jobs/) |2015 |
62 | |  **Agara Labs** | Agara Labs is using deep learning to solve key problems at scale for global enterprises. [Agara Labs Careers](https://agaralabs.com/) | 2017|
63 | |  **Sigmoid** | Specializes in data science and data engineering solutions around the world [Sigmoid Careers](https://www.sigmoid.com/careers/current-openings/) | 2013 |
64 | |  **Tricog** | Tricog has a mission to reduce mortality caused by heart attacks. [Tricog Health Careers](https://angel.co/tricog/jobs)| 2014|
65 | |  **Indegene** | Indegene is an Information Technology and Services company that assists healthcare organisations using analytics and technology. [Indegene Careers](https://careers.indegene.com/careers/Opportunities.aspx)| -- |
66 | |  **Innoplexus** | It aims at cutting drug development costs by mainly using AI. [Innoplexus Careers](https://www.innoplexus.com/index.php/career/) | -- |
67 | |  **HealthReflect** | The platform combines power of machine learning & artificial intelligence (A.I.) to offer predictive & preventive personalised medicine for every patient in India, China, Singapore, UAE & other middle eastern countries. [HealthReflect Careers](https://www.healthreflect.com/career) | -- |
68 | ### Refrences
69 |
70 | * [Towards Data Science](https://towardsdatascience.com/)
71 | * [Data Science Group, IITR](https://medium.com/data-science-group-iitr)
72 | * [data-science-question-answer](https://github.com/ShuaiW/data-science-question-answer#cnn)
73 | * [CS 229 ― Machine Learning CheatSheet](https://stanford.edu/~shervine/teaching/cs-229.html)
74 |
75 | Thank you guys if it helped you kindly share this repo with your fellow developers . Also don't forget to create a pull request if you find something relevant .
76 |
--------------------------------------------------------------------------------
/Reinforcement Learning.md:
--------------------------------------------------------------------------------
1 | # Reinforcement Learning
2 |
--------------------------------------------------------------------------------
/Statistics.md:
--------------------------------------------------------------------------------
1 | # Statistics for Data Science
2 |
3 | **Population** : The entire group one desires information about.
4 | **Sample** : A subset of the population taken because the entire population is usually too large to analyze. It's
5 | characteristics are taken to be representative of the population.
6 | **Mean** : The sum of all the values in the sample divided by the number of values in the sample/population.
7 | **Median** : The median is the value separating the higher half of a data sample from the lower half.
8 | **Standard Deviation** : Square root of the variance. It measures the dispersion around the mean.
9 | **Percentiles** : An extension of median to values other than 50%.
10 | **Interquartile range (IQR)** : the difference between the 75th and 25th percentile
11 | **Mode** : The most frequently occuring value
12 | **Range** : Difference between the maximum value and the minimum value.
13 |
14 | Notice that most of these fall into one of two categories: they capture either the center of the distribution (e.g., mean, median, mode), or its spread (e.g., variance, IQR, range). These two categories are often called **measures of central tendency** and **measures of dispersion**, respectively.
15 |
16 | # Important Distributions
17 |
18 | **Gaussian/Normal** : We say x ∼ N (µ, σ2) to mean that x is drawn from a Gaussian (or Normal) distribution with mean µ and
19 | variance σ2 (or equivalently standard deviation σ). We’ll often use the standard normal distribution, or N (0, 1) (i.e., mean 0 and variance 1).
20 |
21 | * The probability of getting a value within 1 standard deviation of the mean is about 68%. For 2 standard deviations, it’s about 95%, and for 3 standard deviations it’s about 99%. This is sometimes called the **“68-95-99 rule”**.
22 |
23 | **Bernoulli** : A Bernoulli random variable can be thought of as the outcome of flipping a biased coin, where the probability of heads is p. To be more precise, a Bernoulli random variable takes on value 1 with probability p and value 0 with probability 1−p. Its expectation is p, and its variance is p(1 − p).
24 |
25 | Bernoulli variables are typically used to model binary random variables.
26 | ## Central Limit Theorem
27 |
28 | The Central Limit Theorem states that the sampling distribution of the sample means approaches a normal distribution as the sample size gets larger — no matter what the shape of the population distribution. This fact holds especially true for sample sizes over 30.
29 |
30 | ## Confidence Intervals
31 |
32 | A confidence interval is how much uncertainty there is with any particular statistic. Confidence intervals are often used with a margin of error. It tells you how confident you can be that the results from a poll or survey reflect what you would expect to find if it were possible to survey the entire population.
33 |
34 | ## Hypothesis Testing
35 |
36 | Hypothesis testing in statistics is a way for you to test the results of a survey or experiment to see if you have meaningful results.
37 | Hypothesis testing steps:
38 |
39 | * Figure out your null hypothesis,
40 | * State your null hypothesis,
41 | * Choose what kind of test you need to perform,
42 | * Either support or reject the null hypothesis.
43 |
44 | ## Significance Level
45 |
46 | The significance level α is the probability of making the wrong decision when the null hypothesis is true. Alpha levels (sometimes just called “significance levels”) are used in hypothesis tests. Usually, these tests are run with an alpha level of .05 (5%), but other levels commonly used are .01 and .10.
47 |
48 | ## P-value
49 |
50 | A p value is used in hypothesis testing to help you support or reject the null hypothesis. The p value is the evidence against a null hypothesis. The smaller the p-value, the strong the evidence that you should reject the null hypothesis.
51 |
52 | ## A/B testing
53 |
54 | A statistical way of comparing two (or more) techniques, typically an incumbent against a new rival. A/B testing aims to determine not only which technique performs better but also to understand whether the difference is statistically significant. A/B testing usually considers only two techniques using one measurement, but it can be applied to any finite number of techniques and measures.
55 |
56 | ## Correlation
57 |
58 | Correlation is a statistical measure that describes the association between random variables.
59 |
60 | Types of Correlation
61 |
62 | * Pearson Correlation Coefficient (measures the linear association between continuous variables)
63 | * Spearman's Correlation (special case of Pearson ρ applied to ranked (sorted) variables. appropriate to use with both continuous and discrete data.)
64 | * Kendall's Tau (more appropriate for discrete data.)
65 |
66 | ## Statistical Hypothesis Tests :star::star:
67 |
68 | ### Normality Tests (Statistical tests that you can use to check if your data has a Gaussian distribution.)
69 | * **Shapiro-Wilk Test** : Tests whether a data sample has a Gaussian distribution.
70 |
71 | ### Correlation Tests (Statistical tests that you can use to check if two samples are related)
72 |
73 | * **Pearson’s Correlation Coefficient** : Tests whether two samples have a monotonic relationship.
74 |
75 | * **Spearman’s Rank Correlation** : Tests whether two samples have a monotonic relationship.
76 |
77 | * **Chi-Squared Test** : Tests whether two categorical variables are related or independent.
78 |
79 | ### Parametric Statistical Hypothesis Tests (Statistical tests that you can use to compare data samples.)
80 |
81 | * **Student’s t-test** : Tests whether the means of two independent samples are significantly different.
82 |
83 | * **Paired Student’s t-test** : Tests whether the means of two paired samples are significantly different.
84 |
85 | * **Analysis of Variance Test (ANOVA)** : Tests whether the means of two or more independent samples are significantly different.
86 |
87 | * **Repeated Measures ANOVA Test** : Tests whether the means of two or more paired samples are significantly different.
88 |
89 | ### Nonparametric Statistical Hypothesis Tests
90 |
91 | * **Mann-Whitney U Test** : Tests whether the distributions of two independent samples are equal or not.
92 | * **Wilcoxon Signed-Rank Test** : Tests whether the distributions of two paired samples are equal or not.
93 | * **Kruskal-Wallis H Test** : Tests whether the distributions of two or more independent samples are equal or not.
94 | * **Friedman Test** : Tests whether the distributions of two or more paired samples are equal or not.
95 |
96 |
97 |
98 |
--------------------------------------------------------------------------------
/Supervised Learning.md:
--------------------------------------------------------------------------------
1 | # Supervised Learning
2 |
3 | # Table of Contents
4 | * [Supervised Learning](#supervisedlearning)
5 | * [Linear Regression](#linearregression)
6 | * [Logistic Regression](#lr)
7 | * [Multiclass Vs MultiLabel Classification](#mcml)
8 | * [K Nearest Neighbors](#knn)
9 | * [Naive Bayes](#nb)
10 | * [Support Vector Machines](#svm)
11 | * [Bagging](#bagging)
12 | * [Decision Trees](#decision)
13 | * [Random Forest](#rf)
14 | * [Boosting](#boosting)
15 | * [Metrics](#metrics)
16 | ## What is supervised learning?
17 | _Supervised learning_ is where you have input data (X) and their corresponding output variables.
18 | 
19 |
20 | ## What is linear regression ?
21 | _Linear Regression_ is a parametric, discriminative supervised learning algorithm to predict continuous values of a target variable by fitting the best linear relationship between the dependent & independent variable.
22 | 
23 |
24 | ## What is gradient descent ?
25 | _Gradient Descent_ is a first order optimization algorithm which is used for finding the local minima of an
26 | objective function. It starts with intial set of parameter values and iteratively moves towards a set of values that minimize the function. This iterative minimization is done by taking steps towards the negative direction of the function gradient.
27 |
28 | 
29 |
30 | ## What is logistic regression ?
31 | _Logistic regression_ is a parametric, discriminative supervised learning algorithm for classification, i.e used where the response variable is categorical by applying a sigmoid function to a linear prediction.
32 | The idea of logistic regression is to find a relationship between features and probability of particular outcome.
33 |
34 | 
35 |
36 | ## Multiclass Vs Multilabel Classification
37 |
38 | Multiclass classification means a classification task with more than two classes; e.g., classify a set of images of fruits which may be oranges, apples, or pears. Multiclass classification makes the assumption that each sample is assigned to one and only one label: a fruit can be either an apple or a pear but not both at the same time.
39 |
40 | Multilabel classification assigns to each sample a set of target labels. This can be thought as predicting properties of a data-point that are not mutually exclusive, such as topics that are relevant for a document. A text might be about any of religion, politics, finance or education at the same time or none of these.
41 |
42 | ## What is maximum likelihood estimation ?
43 | The principle of maximum likelihood states that we should choose
44 | parameters so as to make the data as high probability as possible. i.e we should choose parameters to maximize likelihood function.
45 |
46 | Note : **Probability** in this mathematical context describes the plausibility of a random outcome, given a model parameter value, without reference to any observed data. **Likelihood** describes the plausibility of a model parameter value, given specific observed data.
47 |
48 | ## What is K- Nearest Neighbors ?
49 | It's a non-parametric supervised learning algorithm in which we assign a label to new data based on the labels of training examples
50 | which are most near to it. It's a lazy learning technique because it goes through complete training data everytime it needs to predict a test sample.
51 |
52 | - Distance Metric
53 | - Euclidean distance
54 | - Manhattan distance
55 |
56 |
57 |
58 | ## How is KNN different from k-means clustering ?
59 | K-Nearest Neighbors is a supervised classication algorithm, while k-means clustering is an unsupervised clustering algorithm. In order for K-Nearest Neighbors to work, you need labeled data you
60 | want to classify an unlabeled point into (thus the nearest neighbor part).
61 | K-means clustering requires only a set of unlabeled points and a
62 | threshold: the algorithm will take unlabeled points and gradually learn how to cluster them into groups by computing the mean of the
63 | distance between different points.
64 |
65 | The critical difference here is that KNN needs labeled points and is thus supervised learning, while k-means doesn’t — and is thus
66 | unsupervised learning.
67 |
68 | ## What is Naive Bayes ?
69 | It is a supervised learning algorithm based on bayes theorem. It classifies
70 | different instances into predefined classes, assuming there is no interdependency of features.
71 |
72 | **Pros**:
73 | - Easy and fast to predict class of test data set. Also, performs well
74 | in multi-class prediction.
75 |
76 | **Cons**:
77 | - Bad estimator: Probability outputs from predict_proba are not to
78 | be taken too seriously.
79 |
80 | - Assumption of independent predictors: In real life, it is almost impossible that we get a set of predictors which are completely independent.
81 |
82 | 
83 |
84 | ## What is Support Vector Machines ?
85 | _Support Vector Machines_ is an non-parametric, discriminative supervised learning algorithm
86 | which identifies optimal separating hyperplane which maximizes the margin between different classes of the training data.
87 |
88 | 
89 |
90 | ### Kernel Functions
91 |
92 | Kernel methods owe their name to the use of kernel functions, which enable them to operate in a high-dimensional feature space without ever computing the coordinates of the data in that space, but rather by simply computing the inner products between the images of all pairs of data in the feature space.
93 |
94 | **Note** : Any linear model can be turned into a non-linear model by applying the kernel trick to the model: replacing its features (predictors) by a kernel function
95 |
96 | 
97 |
98 | ### Examples of SVM Kernels
99 |
100 | * Polynomial kernel
101 | * Gaussian radial basis function (RBF)
102 | * Sigmoid kernel
103 |
104 | **Pros**:
105 |
106 | - It is really effective in higher dimension. If you have more features
107 | than training examples, most of the algorithms perform very bad,
108 | but SVM is the only algorithm which can saves you in this
109 | situation.
110 | - Best algorithm if you data are separable. That two classes are not
111 | mixed.
112 | - Only support vectors affect the optimally spaced hyperplane. So, it
113 | is less affected by outliers.
114 |
115 | **Cons**:
116 |
117 | - On large dataset it takes too much time. Mainly because of kernel
118 | function calculations and finding optimal hyperplane in higher
119 | dimensions.
120 | - Can not perform well in case of overlapping classes.
121 | - Can only give you 0–1 classification. Probably estimates
122 | computation are really expensive.
123 |
124 | ## What is bagging ?
125 | _Bagging_ is an ensemble technique mainly used to reduce the variance of our predictions
126 | by combining the results of multiple classifiers modelled on different sub-samples of the same dataset. In Bagging, individual learner are trained in parallel.
127 |
128 | ## What is decision trees ?
129 | Decision trees are non-parametric supervised learning algorithm.
130 | Given the training data, a decision tree algorithm divides the feature space into regions. For inference, we first see which region does the test data point fall in and take the mean label value (regression) or the majority label value ( classification )
131 |
132 | There are couple of algorithms there to build a decision tree , we only talk about a few which are:
133 |
134 | CART (Classification and Regression Trees) → uses Gini Index(Classification) as metric.
135 | ID3 (Iterative Dichotomiser 3) → uses Entropy function and Information gain as metrics.
136 |
137 | #### Finding the variable/feature for best split.
138 |
139 | **Gini Index**: Measure of variance across all classes of the data. Measures the impurity of the data.
140 | Ex. Given a binary classi cation problem, the number of positive
141 | cases equals the negative ones.
142 | GI = 1/2*(1–1/2)+1/2*(1–1/2) = 1/2
143 |
144 | This is maximum GI possible. As we split data, and move towards subtree, GI decreases to zero with increase in depth of tree.
145 |
146 | **Entropy**: Measure of randomness. More the random data, higher the entropy.
147 | E = -p*log(p) ; p - probability
148 |
149 | **Information Gain**: Decrease in entropy. The difference between the entropy before the split and the average entropy after split is obtained to decide when to split.
150 |
151 | The variable which provides maximum entropy gain is chosen!
152 |
153 | **Pros**:
154 | - Easy to understand and visualise.
155 | - Can be used for feature engineering.
156 | - Very little data preparation needed for algorithm.
157 |
158 | **Cons**:
159 | - If not tuned well, may lead to overfitting.
160 | - Unstable. Small variation in data leads to completely different tree
161 | formation.
162 | - In case of imbalanced dataset, decision trees are biased.However,
163 | by using proper splitting criteria, this issue can be resolved.
164 |
165 | **Important Parameters**:
166 |
167 | - **Minimum Samples Split**: Minimum number of sample required to
168 | split a node. This parameter helps in reducing over tting.
169 | High value: Underfitting, Low value: overfitting.
170 |
171 | - **Maximum Depth of a Tree**: Most in uential parameter. Gives limit
172 | on vertical depth decide upto which level pruning is required.
173 | Higher value: overfitting, Lower value: Underfitting
174 |
175 | - **Maximum Features**: At each node, while splitting either we can
176 | chose best feature from pool of all the features or limited number of
177 | random features. This parameter adds a little randomness - good
178 | generalised model.
179 | 
180 | ## What is random forest ?
181 | Random forest improves bagging further by adding some randomness. In random forest, only a subset of features are selected at random to construct a tree (while often not subsample instances). The benefit is that random forest decorrelates the trees.
182 |
183 | For example, suppose we have a dataset. There is one very predicative feature, and a couple of moderately predicative features. In bagging trees, most of the trees will use this very predicative feature in the top split, and therefore making most of the trees look similar, and highly correlated. Averaging many highly correlated results won't lead to a large reduction in variance compared with uncorrelated results. In random forest for each split we only consider a subset of the features and therefore reduce the variance even further by introducing more uncorrelated trees.
184 |
185 | **Pros**:
186 | - As it predicts by aggregating the predictions from smaller
187 | predictors the variance decreases. Less overfitting.
188 | - Useful when missing data is huge.
189 |
190 | **Cons**:
191 | - Better with classification than regression.
192 | - Black box approach: Many factors are random.
193 | - Slight increase in Bias
194 |
195 | **Parameters**:
196 | - **n_estimators**: Number of trees in the model. The larger the better,
197 | but the longer it will take to compute.
198 |
199 | - **max_features**: Size of the random subsets of features to consider
200 | when splitting a node. Lower the #features, greater the reduction
201 | in variance, but greater the increase in bias.
202 |
203 | - **feature_importances_**: The relative importances of each feature
204 | to the model. Features used in the tree at the top nodes are
205 | relatively more important as more data points are dependent on
206 | that feature.
207 | 
208 |
209 | ## What is boosting ?
210 | Boosting builds on weak learners, and in an iterative fashion. In each iteration, a new learner is added, while all existing learners are kept unchanged. All learners are weighted based on their performance (e.g., accuracy), and after a weak learner is added, the data are re-weighted: examples that are misclassified gain more weights, while examples that are correctly classified lose weights. Thus, future weak learners focus more on examples that previous weak learners misclassified.
211 |
212 | ### Types of Boosting Algorithms
213 |
214 | * **AdaBoost (Adaptive Boosting)**
215 | * **Gradient Tree Boosting**
216 | * **XGBoost**
217 |
218 | 
219 |
220 | **Pros**:
221 |
222 | - Automatically do feature engineering.
223 | - Very little data preparation needed for algorithm.
224 |
225 | **Cons**:
226 |
227 | - Time and computation expensive.
228 | - Complexity of the classiffication increases.
229 | - Hard to implement in real time platform.
230 |
231 |
232 | ## Metrics
233 |
234 | - **Confusion Matrix**
235 |
236 | An NxN matrix where N is the no. of classes, that summarizes how successful a classification model's predictions are.
237 |
238 | - **Accuracy**
239 | Accuracy is the fraction of predictions our model got right.
240 | Suppose you build a model that classified 100 tumors as malignant (the positive class) or benign (the negative class):
241 | 
242 |
243 | - **Recall or Sensitivity or True Positive Rate**
244 | 
245 |
246 | Number of items correctly identified as positive out of total true positives. High recall means you’re not missing many positives.
247 | 
248 | Our model has a recall of 0.11—in other words, it correctly identifies 11% of all malignant tumors.
249 |
250 | - **Precision**
251 | Number of items correctly identified as positive out of total items identified as positive. High precision means low “false alarm rate” (if you test positive, you’re probably positive)
252 | 
253 | Our model has a precision of 0.5—in other words, when it predicts a tumor is malignant, it is correct 50% of the time.
254 | - **True Negative Rate or Specificity**
255 | Number of items correctly identified as negative out of total true negatives.
256 |
257 | - **Type 1 Error or False Positive Rate or false alarm rate**
258 | Number of items wrongly identified as positive out of total true negatives.
259 |
260 | - **Type 2 Error or False Negative Rate or miss rate**
261 | Number of items wrongly identified as negative out of total
262 | true positives.
263 |
264 | - **RMSE (Root Mean Square Error)**
265 | It represents the sample standard deviation of the differences between predicted values and observed values (called residuals).
266 |
267 | - **MAE**
268 | MAE is the average of the absolute difference between the predicted values and observed value.
269 |
270 | - **R Squared (R²) and Adjusted R Squared**
271 |
272 | - R Squared & Adjusted R Squared (_goodness of fit measure_) are often used for explanatory purposes and explains how well your selected independent variable(s) explain the variability in your dependent variable(s).
273 | **Note** : Higher the MSE, smaller the R_squared and poorer is the model.
274 | Just like R², adjusted R² also shows how well terms fit a curve or line but adjusts for the number of terms in a model.
275 | **Note** : The more predictors you add the higher R^2 become hence use adjusted R^2 which adjusts for the degrees of freedom.
276 | 
277 |
278 | ## Explain how a ROC curve works?
279 |
280 | The **ROC curve** is a graphical representation of the contrast between true positive rates and the false positive rate at various thresholds.
281 | It’s often used as a proxy for the trade-off between the sensitivity of the model (true positives) vs the fall-out or the probability it will trigger a false alarm (false positives).
282 |
283 | AUC ROC = area under the ROC curve.
284 |
285 | 
286 |
--------------------------------------------------------------------------------
/Unsupervised Learning.md:
--------------------------------------------------------------------------------
1 | # Unsupervised Learning
2 |
3 | ## What is unsupervised Learning?
4 | Unsupervised learning is where you only have input data (X) and no corresponding output variables.
5 | 
6 |
7 | ## What is clustering?
8 | It's an machine learning technique which segregate the various data points into different groups called clusters such that
9 | entities in a particular group comparatively have more similar traits than entities in another group.
10 |
11 | ## Top 5 Clustering Algorithms to know
12 |
13 | - K-Means Clustering
14 | - Agglomerative Hierarchical Clustering
15 | - Mean-Shift Clustering
16 | - Density-Based Spatial Clustering of Applications with Noise (DBSCAN)
17 | - Expectation–Maximization (EM) Clustering using Gaussian Mixture Models (GMM)
18 |
19 |
20 | ## K- means Clustering ?
21 | k-means clustering aims to partition n observations into k clusters in which each observation belongs to the cluster with the nearest mean.
22 |
23 | 
24 |
25 | ### K-Means Clustering Algorithm
26 |
27 | 
28 |
29 | ## Pros
30 | * K-Means has the advantage that it’s pretty fast, as all we’re really doing is computing the distances between points and group centers; very few computations! It thus has a linear complexity O(n).
31 |
32 | ## Cons
33 | * You have to select how many groups/classes there are.
34 | * K-means also starts with a random choice of cluster centers and therefore it may yield different clustering results on different runs of the algorithm. Thus, the results may not be repeatable and lack consistency.
35 |
36 | ## Determining The Optimal Number Of Clusters: 3 Must Know Methods?
37 | - Elbow method (within-cluster sum of square vs number of clusters) **We want it to be as small as possible**
38 |
39 | 
40 |
41 | - Average silhouette method (average silhouette of observations (avg.sil) vs number of clusters) **high average silhouette width indicates a good clustering**
42 |
43 | 
44 |
45 | - Gap statistic method The gap statistic compares the total intracluster variation for different values of k with their expected values under null reference distribution of the data i.e. a distribution with no obvious clustering.
46 |
47 | 
48 | ## Hierarchical Clustering
49 |
50 | It is a type of connectivity model clustering which is based on the fact that data points that are closer to each other are more similar than the data points lying far away in a data space.
51 |
52 | As the name speaks for itself, the hierarchical clustering forms the hierarchy of the clusters that can be studied by visualising dendogram.
53 |
54 | 
55 |
56 | ## Pros
57 |
58 | * Hierarchical clustering does not require us to specify the number of clusters and we can even select which number of clusters looks best since we are building a tree.
59 |
60 | ## Cons
61 | * Lower efficiency, as it has a time complexity of O(n³)
62 |
63 | ## Dimesionality Reduction Technique
64 |
65 | Mainly there are two types of dimesion reduction technique:
66 | 1. Matrix Factorization(PCA)
67 | 2. Neighbour Graphs (UMAP, T-SNE)
68 |
69 | ### PCA (Principal Component Analysis)
70 | * Principal Component Analysis (PCA) is a dimension reduction technique that projects
71 | the data into a lower dimensional space
72 | * PCA uses Singular Value Decomposition (SVD), which is a matrix factorization method
73 | that decomposes a matrix into three smaller matrices (more details of SVD [here](https://en.wikipedia.org/wiki/Singular-value_decomposition))
74 | * PCA finds top N principal components, which are dimensions along which the data vary
75 | (spread out) the most. Intuitively, the more spread out the data along a specific dimension,
76 | the more information is contained, thus the more important this dimension is for the
77 | pattern recognition of the dataset
78 | * PCA can be used as pre-step for data visualization: reducing high dimensional data
79 | into 2D or 3D.
80 | Note : PCA is interpretable dimension reduction.
81 |
82 | ### T-SNE(t-distributed stochastic neighbor embedding)
83 |
84 | * It is a nonlinear dimensionality reduction technique well-suited for embedding high-dimensional data for visualization in a low-dimensional space of two or three dimensions.
85 | * The t-SNE algorithm comprises two main stages.First, t-SNE constructs a probability distribution over pairs of high-dimensional objects in such a way that similar objects have a high probability of being picked, whilst dissimilar points have an extremely small probability of being picked.
86 | * Second, t-SNE defines a similar probability distribution over the points in the low-dimensional map, and it minimizes the Kullback–Leibler divergence between the two distributions with respect to the locations of the points in the map.
87 |
88 |
89 | ### UMAP(Uniform Manifold Approximation and Projection)
90 |
91 | * UMAP (Uniform Manifold Approximation and Projection) is a novel manifold learning technique for dimension reduction
92 | * The UMAP algorithm is competitive with t-SNE for visualization quality, and arguably preserves more of the global structure with superior run time performance.
93 | * UMAP’s topological foundations allow it to scale to signicantly larger dataset sizes than are feasible for t-SNE.
94 |
--------------------------------------------------------------------------------
/images/10_fold_cv.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/10_fold_cv.png
--------------------------------------------------------------------------------
/images/1_7VyTVnlVj2Ooqa3MRSRFEQ.gif:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/1_7VyTVnlVj2Ooqa3MRSRFEQ.gif
--------------------------------------------------------------------------------
/images/1_ET8kCcPpr893vNZFs8j4xg.gif:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/1_ET8kCcPpr893vNZFs8j4xg.gif
--------------------------------------------------------------------------------
/images/1_KrcZK0xYgTa4qFrVr0fO2w.gif:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/1_KrcZK0xYgTa4qFrVr0fO2w.gif
--------------------------------------------------------------------------------
/images/1_pdfGx9Hfo_01NIiL0A4iPA.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/1_pdfGx9Hfo_01NIiL0A4iPA.png
--------------------------------------------------------------------------------
/images/1_tEo4x2-2iOZcnhSF13rUTA.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/1_tEo4x2-2iOZcnhSF13rUTA.png
--------------------------------------------------------------------------------
/images/20170513_gradient_descent_logistic_animation.gif:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/20170513_gradient_descent_logistic_animation.gif
--------------------------------------------------------------------------------
/images/Adjusted_R2.png_large:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Adjusted_R2.png_large
--------------------------------------------------------------------------------
/images/An-illustration-of-neural-network-with-and-without-dropout.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/An-illustration-of-neural-network-with-and-without-dropout.png
--------------------------------------------------------------------------------
/images/CBOW.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/CBOW.png
--------------------------------------------------------------------------------
/images/Precisionrecall.svg.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Precisionrecall.svg.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-07-03 19-25-29.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-07-03 19-25-29.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-07-08 09-51-27.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-07-08 09-51-27.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-07-09 09-27-32.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-07-09 09-27-32.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-07-09 09-40-16.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-07-09 09-40-16.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-07-09 09-46-53.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-07-09 09-46-53.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-07-09 09-47-10.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-07-09 09-47-10.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-07-10 16-29-57.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-07-10 16-29-57.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-08-12 09-28-06.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-08-12 09-28-06.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-08-21 20-22-44.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-08-21 20-22-44.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-08-21 20-24-08.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-08-21 20-24-08.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-08-21 20-53-48.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-08-21 20-53-48.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-08-21 20-56-24.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-08-21 20-56-24.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-09-21 09-20-55.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-09-21 09-20-55.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-10-04 14-14-08.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-10-04 14-14-08.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-10-04 14-14-25.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-10-04 14-14-25.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-10-04 14-14-52.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-10-04 14-14-52.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-10-04 14-16-00.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-10-04 14-16-00.png
--------------------------------------------------------------------------------
/images/Screenshot from 2018-10-04 14-17-53.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2018-10-04 14-17-53.png
--------------------------------------------------------------------------------
/images/Screenshot from 2019-07-22 19-26-50.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2019-07-22 19-26-50.png
--------------------------------------------------------------------------------
/images/Screenshot from 2019-07-22 19-31-21.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/Screenshot from 2019-07-22 19-31-21.png
--------------------------------------------------------------------------------
/images/WordEmbeddings.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/WordEmbeddings.png
--------------------------------------------------------------------------------
/images/a.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/a.png
--------------------------------------------------------------------------------
/images/b.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/b.png
--------------------------------------------------------------------------------
/images/bag.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/bag.jpg
--------------------------------------------------------------------------------
/images/c.jpeg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/c.jpeg
--------------------------------------------------------------------------------
/images/d.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/d.png
--------------------------------------------------------------------------------
/images/decision-forests-and-discriminant-analysis-77-638.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/decision-forests-and-discriminant-analysis-77-638.jpg
--------------------------------------------------------------------------------
/images/download (1).jpeg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/download (1).jpeg
--------------------------------------------------------------------------------
/images/download.jpeg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/download.jpeg
--------------------------------------------------------------------------------
/images/download.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/download.png
--------------------------------------------------------------------------------
/images/e.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/e.png
--------------------------------------------------------------------------------
/images/edge network.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/edge network.png
--------------------------------------------------------------------------------
/images/f.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/f.png
--------------------------------------------------------------------------------
/images/g.jpeg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/g.jpeg
--------------------------------------------------------------------------------
/images/g.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/g.png
--------------------------------------------------------------------------------
/images/h.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/h.png
--------------------------------------------------------------------------------
/images/healthreflect.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/healthreflect.png
--------------------------------------------------------------------------------
/images/i.jpeg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/i.jpeg
--------------------------------------------------------------------------------
/images/i.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/i.png
--------------------------------------------------------------------------------
/images/image5.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/image5.png
--------------------------------------------------------------------------------
/images/image6-e1536165830511.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/image6-e1536165830511.png
--------------------------------------------------------------------------------
/images/inbox_2181887_2db081ae1af423307f7ae558f76a2063_silhoutte.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/inbox_2181887_2db081ae1af423307f7ae558f76a2063_silhoutte.png
--------------------------------------------------------------------------------
/images/inbox_2181887_373d4f8bda1be2b867c2f404dd335e85_unnamed-chunk-16-1.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/inbox_2181887_373d4f8bda1be2b867c2f404dd335e85_unnamed-chunk-16-1.png
--------------------------------------------------------------------------------
/images/inbox_2181887_42dfac7118608f18d67365a3c35a7aa6_Elbow method.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/inbox_2181887_42dfac7118608f18d67365a3c35a7aa6_Elbow method.png
--------------------------------------------------------------------------------
/images/indigene.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/indigene.png
--------------------------------------------------------------------------------
/images/innoplexus.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/innoplexus.png
--------------------------------------------------------------------------------
/images/j.jpeg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/j.jpeg
--------------------------------------------------------------------------------
/images/kmeansclusteringalgo.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/kmeansclusteringalgo.png
--------------------------------------------------------------------------------
/images/maxresdefault.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/maxresdefault.jpg
--------------------------------------------------------------------------------
/images/me-machine-learning-mathematics-so-true-41965149.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/me-machine-learning-mathematics-so-true-41965149.png
--------------------------------------------------------------------------------
/images/partitions-underfitting-vs-overfitting-regression-via-polynomial-degree.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/partitions-underfitting-vs-overfitting-regression-via-polynomial-degree.png
--------------------------------------------------------------------------------
/images/roc.gif:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/roc.gif
--------------------------------------------------------------------------------
/images/seashell.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/seashell.png
--------------------------------------------------------------------------------
/images/svmhyperplane.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/svmhyperplane.png
--------------------------------------------------------------------------------
/images/workflow.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/theainerd/MLInterview/47d6d3fab31defde30c8ae6a63650a7638d49010/images/workflow.png
--------------------------------------------------------------------------------