Python Deeper Insights into Machine Learning





Course Module 1: Python Machine Learning
Chapter 1: Giving Computers the Ability to Learn from Data 3
Building intelligent machines to transform data into knowledge 4
The three different types of machine learning 4
An introduction to the basic terminology and notations 10
A roadmap for building machine learning systems 12
Using Python for machine learning 15
Summary 17
Chapter 2: Training Machine Learning Algorithms
for Classification 19
Artificial neurons – a brief glimpse into the early history of
machine learning 20
Implementing a perceptron learning algorithm in Python 26
Adaptive linear neurons and the convergence of learning 35
Summary 49
Chapter 3: A Tour of Machine Learning Classifiers
Using Scikit-learn 51
Choosing a classification algorithm 51
First steps with scikit-learn 52
Modeling class probabilities via logistic regression 58
Maximum margin classification with support vector machines 71
Solving nonlinear problems using a kernel SVM 77
Decision tree learning 82
Table of Contents
[ ii ]
K-nearest neighbors – a lazy learning algorithm 94
Summary 98
Chapter 4: Building Good Training Sets – Data Preprocessing 101
Dealing with missing data 101
Handling categorical data 106
Partitioning a dataset in training and test sets 110
Bringing features onto the same scale 112
Selecting meaningful features 114
Assessing feature importance with random forests 126
Summary 128
Chapter 5: Compressing Data via Dimensionality Reduction 129
Unsupervised dimensionality reduction via principal
component analysis 130
Supervised data compression via linear discriminant analysis 140
Using kernel principal component analysis for nonlinear mappings 150
Summary 169
Chapter 6: Learning Best Practices for Model Evaluation
and Hyperparameter Tuning 171
Streamlining workflows with pipelines 171
Using k-fold cross-validation to assess model performance 175
Debugging algorithms with learning and validation curves 181
Fine-tuning machine learning models via grid search 187
Looking at different performance evaluation metrics 191
Summary 200
Chapter 7: Combining Different Models for Ensemble Learning 201
Learning with ensembles 201
Implementing a simple majority vote classifier 205
Evaluating and tuning the ensemble classifier 215
Bagging – building an ensemble of classifiers from
bootstrap samples 221
Leveraging weak learners via adaptive boosting 226
Summary 234
Chapter 8: Applying Machine Learning to Sentiment Analysis 235
Obtaining the IMDb movie review dataset 235
Introducing the bag-of-words model 238
Training a logistic regression model for document classification 246
Working with bigger data – online algorithms and out-of-core learning 248
Summary 252
Table of Contents
[ iii ]
Chapter 9: Embedding a Machine Learning Model into a
Web Application 253
Serializing fitted scikit-learn estimators 254
Setting up a SQLite database for data storage 257
Developing a web application with Flask 259
Turning the movie classifier into a web application 266
Deploying the web application to a public server 274
Summary 278
Chapter 10: Predicting Continuous Target Variables with
Regression Analysis 279
Introducing a simple linear regression model 280
Exploring the Housing Dataset 281
Implementing an ordinary least squares linear regression model 287
Fitting a robust regression model using RANSAC 293
Evaluating the performance of linear regression models 296
Using regularized methods for regression 299
Turning a linear regression model into a curve – polynomial
regression 300
Summary 311
Chapter 11: Working with Unlabeled Data – Clustering Analysis 313
Grouping objects by similarity using k-means 314
Organizing clusters as a hierarchical tree 328
Locating regions of high density via DBSCAN 336
Summary 342
Chapter 12: Training Artificial Neural Networks for
Image Recognition 343
Modeling complex functions with artificial neural networks 344
Classifying handwritten digits 352
Training an artificial neural network 367
Developing your intuition for backpropagation 374
Debugging neural networks with gradient checking 375
Convergence in neural networks 381
Other neural network architectures 383
A few last words about neural network implementation 386
Summary 387
Table of Contents
[ iv ]
Chapter 13: Parallelizing Neural Network Training with Theano 389
Building, compiling, and running expressions with Theano 390
Choosing activation functions for feedforward neural networks 403
Training neural networks efficiently using Keras 410
Summary 416
Course Module 2: Designing Machine Learning
Systems with Python
Chapter 1: Thinking in Machine Learning 421
The human interface 422
Design principles 425
Summary 453
Chapter 2: Tools and Techniques 455
Python for machine learning 456
IPython console 456
Installing the SciPy stack 457
NumPY 458
Matplotlib 464
Pandas 468
SciPy 471
Scikit-learn 474
Summary 481
Chapter 3: Turning Data into Information 483
What is data? 484
Big data 484
Signals 500
Cleaning data 502
Visualizing data 504
Summary 507
Chapter 4: Models – Learning from Information 509
Logical models 509
Tree models 517
Rule models 521
Summary 528
Table of Contents
[ v ]
Chapter 5: Linear Models 529
Introducing least squares 530
Logistic regression 538
Multiclass classification 544
Regularization 545
Summary 548
Chapter 6: Neural Networks 549
Getting started with neural networks 549
Logistic units 551
Cost function 556
Implementing a neural network 559
Gradient checking 565
Other neural net architectures 566
Summary 567
Chapter 7: Features – How Algorithms See the World 569
Feature types 570
Operations and statistics 571
Structured features 574
Transforming features 574
Principle component analysis 583
Summary 585
Chapter 8: Learning with Ensembles 587
Ensemble types 587
Bagging 588
Boosting 594
Ensemble strategies 601
Summary 604
Chapter 9: Design Strategies and Case Studies 605
Evaluating model performance 605
Model selection 610
Learning curves 613
Real-world case studies 615
Machine learning at a glance 626
Summary 627
Table of Contents
[ vi ]
Course Module 3: Advanced Machine
Learning with Python
Chapter 1: Unsupervised Machine Learning 631
Principal component analysis 632
Introducing k-means clustering 637
Self-organizing maps 648
Further reading 654
Summary 655
Chapter 2: Deep Belief Networks 657
Neural networks – a primer 658
Restricted Boltzmann Machine 663
Deep belief networks 679
Further reading 685
Summary 686
Chapter 3: Stacked Denoising Autoencoders 687
Autoencoders 687
Stacked Denoising Autoencoders 696
Further reading 705
Summary 705
Chapter 4: Convolutional Neural Networks 707
Introducing the CNN 707
Further Reading 729
Summary 730
Chapter 5: Semi-Supervised Learning 731
Introduction 731
Understanding semi-supervised learning 732
Semi-supervised algorithms in action 733
Further reading 756
Summary 757
Chapter 6: Text Feature Engineering 759
Introduction 759
Text feature engineering 760
Further reading 783
Summary 784
Chapter 7: Feature Engineering Part II 785
Introduction 785
Creating a feature set 786
Table of Contents
[ vii ]
Feature engineering in practice 805
Further reading 829
Summary 830
Chapter 8: Ensemble Methods 831
Introducing ensembles 832
Using models in dynamic applications 851
Further reading 863
Summary 864
Chapter 9: Additional Python Machine Learning Tools 865
Alternative development tools 866
Further reading 875
Summary 875
Chapter 10: Chapter Code Requirements 879
Biblography 881