Better Deep Learning _Jason Brownlee.pdf

Better Deep Learning Train Faster, Reduce Overfitting, and Make Better Predictions Jason Brownlee i Disclaimer The information contained within this eBook is strictly for educational purposes. If you wish to apply ideas contained in this eBook, you are taking full responsibility for your actions. The author has made every eort to ensure the accuracy of the information within this book was correct at time of publication. The author does not assume and hereby disclaims any liability to any party for any loss, damage, or disruption caused by errors or omissions, whether such errors or omissions result from accident, negligence, or any other cause. No part of this eBook may be reproduced or transmitted in any form or by any means, electronic or mechanical, recording or by any information storage and retrieval system, without written permission from the author. Acknowledgements Special thanks to my proofreader Sarah Martin and my technical editors Andrei Cheremskoy, Michael Sanderson, Arun Koshy. Copyright Better Deep Learning ©Copyright 2019 Jason Brownlee. All Rights Reserved. Edition: v1.3 Contents Copyright i Contents ii Preface iii Introduction Welcome v Framework for Better Deep Learning Diagnostic Learning Curves I Better Learning 1 Improve Learning by Understanding Optimization 1.1 Neural Nets Learn a Mapping Function 1.2 Learning Network Weights Is Hard 1.3 Key Features of the Error Surface 1.4 Navigating the Non-Convex Error Surface 1.5 Implications for Training 1.6 Components of the Learning Algorithm 1.7 Further Reading 1.8 Summary 2 Congure Capacity with Nodes and Layers 2.1 Neural Network Model Capacity 2.2 Nodes and Layers Keras API 2.3 Model Capacity Case Study 2.4 Extensions 2.5 Further Reading 2.6 Summary ii CONTENTS iii 3 Congure Gradient Precision with Batch Size 3.1 Batch Size and Gradient Descent 3.2 Gradient Descent Keras API 3.3 Batch Size Case Study 3.4 Extensions 3.5 Further Reading 3.6 Summary 4 Congure What to Optimize with Loss Functi