Adaptive Machine Learning Algorithms with Python: Solve Data Analytics and Machine Learning Problems on Edge Devices

ISBN-13: 9781484280164 / Angielski / Miękka / 2022 / 300 str.

Chanchal Chatterjee;Vwani Roy Chowdhury

Opis

Learn to use adaptive algorithms to solve real-world streaming data problems. This book covers a multitude of data processing challenges, ranging from the simple to the complex. At each step, you will gain insight into real-world use cases, find solutions, explore code used to solve these problems, and create new algorithms for your own use.Authors Chanchal Chatterjee and Vwani P. Roychowdhury begin by introducing a common framework for creating adaptive algorithms, and demonstrating how to use it to address various streaming data issues. Examples range from using matrix functions to solve machine learning and data analysis problems to more critical edge computation problems. They handle time-varying, non-stationary data with minimal compute, memory, latency, and bandwidth. Upon finishing this book, you will have a solid understanding of how to solve adaptive machine learning and data analytics problems and be able to derive new algorithms for your own use cases. You will also come away with solutions to high volume time-varying data with high dimensionality in a low compute, low latency environment.What You Will Learn

• Apply adaptive algorithms to practical applications and examples
• Understand the relevant data representation features and computational models for time-varying multi-dimensional data
• Derive adaptive algorithms for mean, median, covariance, eigenvectors (PCA) and generalized eigenvectors with experiments on real data
• Speed up your algorithms and put them to use on real-world stationary and non-stationary data
• Master the applications of adaptive algorithms on critical edge device computation applications

Who This Book Is For

Machine learning engineers, data scientist and architects, software engineers and architects handling edge device computation and data management.

Spis treści

​Chapter 1. Introducing Data Representation Features

Set the context for the reader with important data representation features, present the need for adaptive algorithms to compute them and demonstrate how these algorithms are important in multiple disciplines. Additionally, discuss a common methodology adopted to derive all our algorithms.

Sub-topics:

1. Data representation features

2. Computational models for time-varying multi-dimensional data

3. Multi-disciplinary origin of adaptive algorithms

4. Common Methodology for Derivations of Algorithms

5. Outline of The Book

Chapter 2. General Theories and Notations

Introduce the reader to types of data in real-world streaming applications, discuss practical use cases and derive adaptive algorithms for mean, normalized mean, median, and covariances. Support the results with experiments on real data.

Sub-topics:

1. Introduction

2. Stationary and Non-Stationary Sequences

3. Use Cases for Algorithms Covered in this Chapter 

4. Adaptive Mean and Covariance of Nonstationary Sequences

5. Adaptive Covariance and Inverses

6. Adaptive Normalized Mean Algorithm

7. Adaptive Median Algorithm

8. Experimental Results

Chapter 3. Square Root and Inverse Square Root

Introduce readers to practical applications of square roots and inverse square roots of streaming data matrices, then present algorithms to compute them. Support the algorithms with real data.

Sub-topics:

1. Introduction and Use Cases

2. Adaptive Square Root Algorithms

3. Adaptive Inverse Square Root Algorithms

4. Experimental Results

Chapter 4. First Principal Eigenvector

Introduce the reader to adaptive computation of first principal component of streaming data, discuss the use cases with examples, derive ten algorithms with the common methodology adopted here. Demonstrate the algorithms with real-world non-stationary streaming data examples.

Sub-topics:

1. Introduction and Use Cases

2. Algorithms and Objective Functions

3. OJA Algorithm

4. RQ, OJAN, and LUO Algorithms

5. IT and XU Algorithms

6. Penalty Function Algorithm

7. Augmented Lagrangian Algorithms

8. Summary of Algorithms

9. Experimental Results

Chapter 5. Principal and Minor Eigenvectors

Introduce the reader to adaptive computation of all principal components, discuss powerful use cases with examples, derive 21 adaptive algorithms and demonstrate the algorithms on real-world time-varying data.

Sub-topics:

1. Introduction and Use Cases

2. Algorithms and Objective Functions

3. OJA Algorithms

4. XU Algorithms

5. PF Algorithms

6. AL1 Algorithms

7. AL2 Algorithms

8. IT Algorithms

9. RQ Algorithms

10. Summary of Adaptive Eigenvector Algorithms

11. Experimental Results

Chapter 6. Accelerated Computation eigenvectors

Introduce the reader to methods to speed up the adaptive algorithms presented in this book. Help the reader speed up a few algorithms and demonstrate their usefulness and acceleration on real-world stationery and non-stationary data.

Sub-topics:

1. Introduction

2. Gradient Descent Algorithm

3. Steepest Descent Algorithm

4. Conjugate Direction Algorithm

5. Newton-Raphson Algorithm

6. Experimental Results

Chapter 7. Generalized Eigenvectors

Introduce the reader to the adaptive computation of generalized eigenvectors of streaming data matrices in real-time applications. Discuss use cases and algorithms and show experimental results on real data.

Sub-topics:

1. Introduction and Use Cases

2. Algorithms and Objective Functions

3. OJA GEVD Algorithms

4. XU GEVD Algorithms

5. PF GEVD Algorithms

6. AL1 GEVD Algorithms

7. AL2 GEVD Algorithms

8. IT GEVD Algorithms

9. RQ GEVD Algorithms

10. Experimental Results

Chapter 8. Real–World Applications Linear Algorithms

Help the reader understand real-world applications of the adaptive algorithms. Demonstrate five important applications of adaptive algorithms on critical edge device computation applications.

Sub-topics:

1. Detecting Feature Drift

2. Adapt to Incoming Data Drift

3. Compress High Volume Data

4. Detecting Feature Anomalies

O autorze

Chanchal Chatterjee, Ph.D, has held several leadership roles in machine learning, deep learning and real-time analytics. He is currently leading Machine Learning and Artificial Intelligence at Google Cloud Platform, California, USA. Previously, he was the Chief Architect of EMC CTO Office where he led end-to-end deep learning and machine learning solutions for data centers, smart buildings, and smart manufacturing for leading customers. Chanchal received several awards including an Outstanding paper award from IEEE Neural Network Council for adaptive learning algorithms recommended by MIT professor Marvin Minsky. Chanchal founded two tech startups between 2008-2013. Chanchal has 29 granted or pending patents, and over 30 publications. Chanchal received M.S. and Ph.D. degrees in Electrical and Computer Engineering from Purdue University.

Blurb

Learn to use adaptive algorithms to solve real-world streaming data problems. This book covers a multitude of data processing challenges, ranging from the simple to the complex. At each step, you will gain insight into real-world use cases, find solutions, explore code used to solve these problems, and create new algorithms for your own use.

Authors Chanchal Chatterjee and Vwani P. Roychowdhury begin by introducing a common framework for creating adaptive algorithms, and demonstrating how to use it to address various streaming data issues. Examples range from using matrix functions to solve machine learning and data analysis problems to more critical edge computation problems. They handle time-varying, non-stationary data with minimal compute, memory, latency, and bandwidth. 

Upon finishing this book, you will have a solid understanding of how to solve adaptive machine learning and data analytics problems and be able to derive new algorithms for your own use cases. You will also come away with solutions to high volume time-varying data with high dimensionality in a low compute, low latency environment.

You will:

• Apply adaptive algorithms to practical applications and examples
• Understand the relevant data representation features and computational models for time-varying multi-dimensional data
• Derive adaptive algorithms for mean, median, covariance, eigenvectors (PCA) and generalized eigenvectors with experiments on real data
• Speed up your algorithms and put them to use on real-world stationary and non-stationary data
• Master the applications of adaptive algorithms on critical edge device computation applications