5. Dataset and subsets:

 

Training Set:

·       Training set is used for learning the parameters of the model.

·       Training set has an optimistic bias.

·       The training set is usually the biggest one of three sets and used to build the model.

Test Set:

·       Test set is used to get a final, unbiased estimate of how well the learning method works.

·       We expect that this estimate to be worse than training/validation set.

·       Test set is not biased and this distinguishes test set from training set.

·       Similar to Training set, Test set is a finite sample and bound to have variance due to sample size, but the test set doesn’t have an optimistic or pessimistic bias.

·       The test set doesn’t effect the outcome of learning process which only uses training set.

Validation Set:

·       Validation set is not used for learning but is for avoiding overfitting.

·       The idea of ‘Validation Set’ is identical to ‘Test Set’ but there is a difference.

·       We remove a subset from data and this subset is not used for training.

·       Although validation set is not directly used for training, it will be used in making certain choices in the learning process. As the set affects the learning process, it is no longer a test set.

·       The more we use validation set to fine tune the model, the more validation set becomes like a training set.

·       The validation and test sets are roughly the same sizes, much smaller than the size of the training set. The learning algorithm cannot use examples from these two subsets to build the model.

·       Validation set is used for selecting the model complexity.

In the past, the rule of thumb was to use 70% of the dataset for training, 15% for validation and 15% for testing. However, in the age of big data, datasets often have millions of examples. In such cases, it could be reasonable to keep 95% for training and 2.5%/2.5% for validation/testing.

 

4. MACHINE LEARNING TOOLS:

 

3. MACHINE LEARNING FRAMEWORK:

 ·       y = f(x)

o   y = output, x = feature representation, f( ) is prediction function

·       Training: Given a training set, estimate the prediction function f( ) by minimizing the prediction error.

·       Testing: Apply f ( ) to unknown test sample x and predicted value (output) is y.

·       y = f(w, x) (for linear model)

o   y = output, x = feature representation, w = weight, f(w, ) = prediction function

·       Training: Given a training set, estimate the prediction function, f( ) by minimizing the prediction error.

·       Testing: Apply f(w, ) to unknown test sample x and predicted value (output) is y

o   Parameter: primary problem is to find the parameters W.

AI/ML Problem – phases:

The following are the various steps involved in problem solving using AI/ML:

·       Define your task

·       Collect Data

·       Preprocessing of Data

·       Dimensionality Reduction/Feature Selection

·       Choose ML Algorithm

·       Experimental Design

·       Test and Validate

·       Run System

MACHINE LEARNING WORK FLOW:

2. MACHINE LEARNING - INTRODUCTION

 DEFINITION:

Machine Learning is a set of methods that can automatically detect patterns in data and then use the uncovered patterns to predict future data or to perform other kinds of decision making under uncertainty (such as planning how to collect more data).

LEARNING PROBLEM:

A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.

Example: a computer program that learns to play checkers P in this case is measured by its ability to win

T in this case is playing checkers games

E in this case is obtained by playing games against itself

In general, to have a well-defined learning problem we must identify these three features:

1.         The class of tasks (T)

2.         The measure of performance to be improved (P)

3.         The source of experience (E)

Checkers learning problem:

·       Task T – playing checkers

·       Performance measure P – percent of games won against opponents

·       Training experience E – playing practice games against itself

 

Handwriting recognition learning problem:

·       Task T – recognizing and classifying handwritten words within images

·       Performance measure P – percentage of words correctly classified

·       Training Experience E – a database of handwritten words with given classifications

Robot driving learning problem:

·       Task T – driving on public four lane highways using vision sensors

·       Performance measure P – average distance travelled before an error (as judged by human overseer)

·       Training experience E – a sequence of images and steering commands recorded while observing a human driver

E-mail classification:

·       Task T: Categorize email messages as spam or legitimate.

·       Performance P: Percentage of email messages correctly classified.

·       Training Experience E: Database of emails, some with human-given labels

1.Machine Learning – TABLE OF CONTENTS

       1.         Machine Learning – Introduction

2.         Machine Learning Framework

3.         AI/ML Problem – phases

4.         ML Workflow, Tools and Landscape

5.         Dataset and subsets

6.         Machine Learning Types

7.         Supervised Learning

8.         Supervised Learning Types

9.         Classification Vs Regression

10.     Performance metrics

11.     Classification Algorithms

a.      Logistic Regression

b.      K-Nearest Neighbor (KNN)

c.      Decision Tree

d.      Random Forest – Gradient Boosting

e.      Naïve Bayes

f.       Support Vector Machines (SVM)

g.      Gradient Descent

h.      Neural Networks

12.     Neural Networks

a.      Perceptron

b.      Multi-Level Perceptron (MLP)

c.      Artificial Neural Networks (ANN)

d.      Deep Neural Networks (DNN)

13.     Activation Functions

14.     Dropout

15.     Backpropagation

16.     Cross Validation

17.     Model Compression

18.     Loss Function

19.     Deep Learning/Deep Neural Networks

a.      Convolutional Neural networks (CNN)

b.      Recurrent Neural Networks (RNN)

c.      Restricted Boltzmann Machine (RBM)

d.      Deep Belief Networks (DBN)

e.      Autoencoders

20.     Convolutional Neural Networks

a.      Core idea

b.      Convolution layer

c.      Various terms associated with CNN

d.      Pooling layer and types of pooling

e.      CNN Architecture

f.       Optimization of CNN

g.      Working of CNN

h.      Applications

21.     Recurrent Neural Networks

a.      RNN and limitations

b.      Differences with CNN

c.      Long Short-Term Memory (LSTM)

d.      Gated Feedback Recurrent Neural Networks (GRU)

22.     Restricted Boltzmann Machine (RBM)

23.     Deep Belief Networks (DBN)

24.     Autoencoders

a.      Types

b.      Applications

c.      Convolutional Autoencoders

d.      Denoising Autoencoders

e.      Deep Autoencoders

f.       Variational Autoencoders (VAE)

g.      Sparse Autoencoders

25.     Siamese Neural Networks

26.     Generative Adversarial Networks (GAN)

27.     Regression Algorithms

a.      Linear Regression

b.      Multiple Linear Regression

c.      Polynomial Regression

d.      Ridge Regression

e.      Lasso Regression

f.       ElasticNet Regression

28.     Unsupervised Learning

29.     Unsupervised Learning Algorithms

a.      Clustering

b.      K-Means Clustering

c.      Hierarchical Clustering

d.      DBSCAN

e.      Gaussian Mixtures

f.       Spectral Clustering

30.     Dimensionality Reduction

a.      Feature Engineering

b.      Multicollinearity

c.      Factor Analysis

d.      Principal Component Analysis (PCA)

e.      Linear Discriminant Analysis (LDA)

f.       Isometric Mapping (IsoMap)

g.      Locally Linear Embedding (LLE)

h.      t-distribution Stochastic Neighborhood Embedding (t-SNE)

31.     Semi-supervised Learning

32.     Semi-supervised learning algorithms:

a.      Self Training

b.      Generative Models

c.      S3VMs

d.      Graph Based Algorithms

e.      Multiview algorithms

33.     Reinforcement Learning

a.      Introduction

b.      Approaches

c.      Markov Decision Process

d.      Q-learning

e.      Application of RL

34.     Other topics

a.      Word2Vec

b.      Generalization

c.      Overfitting

d.      Regularization

e.      Bias

f.       Variance

g.      Occam’s Razor

h.      Bag of Words (BoW)

i.       Recommender Systems

j.       Ensemble Methods

k.      Natural Language Processing (NLP)