Deep Learning with C#, .Net and Kelp.Net

Cole Matt R.

Description

Get hands on with Kelp.Net, Microsoft's latest Deep Learning frameworkKey features Deep Learning Basics The ultimate Kelp.Net reference guide Develop state of the art deep learning applications C# deep learning code Develop advanced deep learning models with minimal code Develop your own advanced deep learning models Loading and Saving Deep Learning Models Comprehensive Kelp.Net reference Sample Deep Learning Models and Tests penCL Reference Easily add deep learning to your applications Many sample models and tests Intuitive and user friendly Description Deep Learning with Kelp.Net is the ultimate reference for C# .Net developers who are passionate about deep learning. Readers will learn all the skills necessary to develop powerful, scalable and flexible deep learning models from a fluid and easy to use API. Upon completing the book the reader will have all the tools necessary to add powerful deep learning capabilities to their new or existing applications.What will you learn In-depth knowledge of Kelp.Net How to develop deep learning models C# deep learning programming Open-Computing Language (OpenCL) Loading and saving deep learning models How to develop and use activation functions How to test deep learning modelsWho this book is for This book targets C# .Net developers who are passionate about deep learning yet want to do so from an easy and intuitive API.Table of contents1. Introduction2. ML/DL Terms and Concepts3. Deep Instrumentation4. Kelp.Net Reference5. Loading and Saving Models6. Model Testing and Training7. Sample Deep Learning Tests8. Creating Your Own Deep Learning Tests9. Appendix A: Evaluation Metrics10. Appendix B: OpenCL About the authorMatt R. Cole is a seasoned developer and published author with over 30 years' experience in Microsoft Windows, C, C++, C# and .Net. Matt is the owner of Evolved AI Solutions, a premier provider of advanced Machine Learning/Bio-AI technologies. Matt developed the first enterprise grade MicroService framework written completely in C# and .Net, which is used in production by a major hedge fund in NYC. Matt also developed the first Bio Artificial Intelligence framework which completely integrates mirror and canonical neurons. He continues to push the limits of Machine Learning, Biological Artificial Intelligence, Deep Learning and MicroServices. In his spare time Matt loves to continue his education and contribute to open source efforts such as Kelp.Net. His Website: www.evolvedaisolutions.comHis LinkedIn Profile: https://www.linkedin.com/in/evolvedai/His Blog: https://evolvedaisolutions.com/blog.html

Sujets

• Informatique
• Computer science

Informations

• Publié par: BPB Publications
• Date de parution: 20 septembre 2019
• Nombre de lectures: 0
• EAN13: 9789389423747
• Langue: English
• Poids de l'ouvrage: 1 Mo

Extrait

Deep Learning with C#, .NET and Kelp.NET
The Ultimate Kelp.Net Deep Learning Guide
by Matt R. Cole
FIRST EDITION 2019
Copyright © BPB Publications, India
ISBN: 978-93-88511-018
All Rights Reserved. No part of this publication may be reproduced or distributed in any form or by any means or stored in a database or retrieval system, without the prior written permission of the publisher with the exception to the program listings which may be entered, stored and executed in a computer system, but they can not be reproduced by the means of publication.
LIMITS OF LIABILITY AND DISCLAIMER OF WARRANTY
The information contained in this book is true to correct and the best of author’s & publisher’s knowledge. The author has made every effort to ensure the accuracy of these publications, but cannot be held responsible for any loss or damage arising from any information in this book.
All trademarks referred to in the book are acknowledged as properties of their respective owners.
 
 
Distributors:
BPB PUBLICATIONS 20, Ansari Road, Darya Ganj New Delhi-110002 Ph: 23254990/23254991
MICRO MEDIA Shop No. 5, Mahendra Chambers, 150 DN Rd. Next to Capital Cinema, V.T. (C.S.T.) Station, MUMBAI-400 001 Ph: 22078296/22078297
DECCAN AGENCIES 4-3-329, Bank Street, Hyderabad-500195 Ph: 24756967/24756400
BPB BOOK CENTRE 376 Old Lajpat Rai Market, Delhi-110006 Ph: 23861747
 
 
Published by Manish Jain for BPB Publications, 20, Ansari Road, Darya Ganj, New Delhi-110002 and Printed by Repro India Pvt Ltd, Mumbai
About the Author
Matt R. Cole is a seasoned developer and author with over 30 years’ experience in Microsoft Windows, C, C++, C# and .Net. He is the owner of Evolved AI Solutions, a premier provider of advanced Machine Learning/Bio-AI technologies. He developed the first enterprise-grade microservice framework (written completely in C# and .Net) used by a major hedge fund in NYC and also developed the first Bio Artificial Intelligence Swarm framework which completely integrates mirror and canonical neurons. He continues to push the limits of Machine Learning, Biological or Swarm Artificial Intelligence, Deep Learning and MicroServices. In his spare time, he continues his education taking every available course in advanced Mathematics, AL/ML/DL, Quantum Mechanics/Physics, String Theory and Computational Neuroscience and contributes to open source efforts such as Kelp.Net.
Reviewer
Gaurav Arora has done M.Phil in computer science. He is a Microsoft X-MVP, life time member of Computer Society of India (CSI), Advisory member of IndiaMentor, certified as a scrum trainer/coach, XEN for ITIL-F and APMG for PRINCE-F and PRINCE-P. He is an Open source developer, contributor to TechNet Wiki, Founder of Ovatic Systems Private Limited. In 22+ years of his career, he has mentored thousands of students and industry professionals. You can tweet Gaurav on his twitter handle @g_arora.
Preface
For those of you who are C# developers, you know how painfully hard it is to find good examples of how to implement deep learning in your applications without resorting to using languages such as Python and R. This book is designed to give you all you need to accomplish this goal. You will find that Kelp.Net is an invaluable tool used to create powerful and expressive deep learning models. But I know that along the way, someone might ask you what is happening behind the scenes. This is why I have also chosen to include and deeply integrate ReflectInsight into this book, so that all the output goes into this power logging application. One of the hardest questions to answer is what your deep learning models are doing behind the scenes. This book makes it easy to answer this question. Not only will you be able to easily create your own deep learning models, but you will also gain a better understanding about what is going on behind the scenes. This book is a must for every C# developer who wants to learn deep learning and is the penultimate reference guide for Kelp. Net.
Acknowledgements
This book would not have been possible without the incredible support I received from my wife, Neda every day. I would also like to thank everyone at BPB Publications for giving me the opportunity to write this incredible book and for being an incredible supportive platform for authors. I would like to thank Mindy and Cocoa for their unique kind of support. I wouldn’t have it any other way! And finally, a big thank you all the readers for purchasing this book and taking your first (or next) step down the road to deep learning. May your journey be a good one!
Downloading the code bundle and colored images:
 
Please follow the link to download the Code Bundle and the Colored Images of the book:
https://rebrand.ly/50f4c
 
Errata
We take immense pride in our work at BPB Publications and follow best practices to ensure the accuracy of our content to provide with an indulging reading experience to our subscribers. Our readers are our mirrors, and we use their inputs to reflect and improve upon human errors if any, occurred during the publishing processes involved. To let us maintain the quality and help us reach out to any readers who might be having difficulties due to any unforeseen errors, please write to us at :
[e-mail]
Your support, suggestions and feedbacks are highly appreciated by the BPB Publications’ Family.
Table of Contents
1. Take This ___ and ___ It
Objectives of this book
Neural network overview
Machine learning overview
Deep learning overview
Complexity
Machine and deep learning differences
Summary
2. Machine Learning/Deep Learning Terms and Concepts
Overview
Neuron/Perceptron
Multi-Layer Perceptron (MLP)
Features
Weights
Bias
Activation Function
Sigmoid
ReLU (Rectified Linear Units)
Softmax
Neural network
Input/Output/Hidden Layers
Forward propagation
Back propagation
The No Free Lunch theorem
The Curse of Dimensionality
The more neurons versus more layers
Cost function
Gradient descent
Learning rate
Batches/Batch size
Epochs
Iterations
Dropout
Batch Normalization
CNN (Convolutional Neural Network)
Pooling
Padding
Recurrent neuron
RNN (Recurrent Neural Network)
Vanishing gradient problem
Exploding gradient problem
Logistic Neurons
Hidden layers
Types of neural networks
Generalization
Regularization
Loss
Loss over time
Loss versus learning curve
Supervised learning
Bias-Variance Trade-off (overfitting and underfitting)
Bias
Variance
Overfitting
Is your model overfitting or underfitting?
Prevention of overfitting and underfitting
Amount of training data
Input space dimensionality
Incorrect output values
Data heterogeneity
Unsupervised learning
Reinforcement learning
Manifold learning
Types of manifolds in deep learning
Topological
Differentiable
Riemannian
Principal Component Analysis (PCA)
Hyperparameter training
Approaches to hyperparameter tuning
Grid search
Random search
Bayesian optimization
Gradient-based optimization
Evolutionary optimization
Summary
References
3. Deep Instrumentation Using ReflectInsight
Next generation logging viewers
Message log
Message details
Message properties
Bookmarks
Call Stack
Message Navigation
Advanced Search
User-Defined Views and Filtering
Auto Save/Purge rolling log files
Watches
Time zone formatting
Router
Log viewer
Live viewer
SDK
Configuration editor
Overview
XML configuration
Dynamic configuration
Configuration editor
Message type logging reference
Assertions
Assigned variables
Attachments
Audit failure and success
Checkmarks
Checkpoints
Collections
Comments
Currency
Data
DataSet
DataSetSchema
DataTable
DataTableSchema
DataView
Date/Time
Debug
Desktop Image
Errors
Exceptions
Fatal Errors
Generations
Images
Information
Levels
Linq queries and results
Loaded assemblies
Loaded processes
Memory status
Messages
Notes
Process Information
Reminders
Serialized Objects
SQL strings
Stack Traces
System Information
Text files
Thread Information
Typed collections
Warning
XML
XML files
Tracing method calls
Attaching message properties
To one request
To all requests
To a single message
Watches
Using custom data
Output
Summary
4. Kelp.Net Reference
Let us be honest
Downloading Kelp.Net
Building the source code
What is Kelp.Net?
N-dimensional arrays
Optimizers
AdaDelta
AdaGrad
Adam
GradientClippin g
MomentumSGD
RMSprop
SGD
Poolings
MaxPooling
AveragePooling
FunctionStack
FunctionDictionary
SplitFunction
SortedList
SortedFunctionStack
Activation Functions
Activation plots
ArcSinH
ArcTan
ELU
Gaussian
LeakyReLU
LeakyReLUShifted
LogisticFunction
MaxMinusOne
PolynomialApproximantSteep
QuadraticSigmoid
RbfGaussian
ReLU
ReLuTanh
ScaledELU
Sigmoid
Sine
Softmax
Softplus
SReLU
SReLUShifted
Swish
Tanh
Connections
Convolution2D
Deconvolution2D
EmbedID
Linear
LSTM
Normalization
BatchNormalization
Local Response Normalization
Noise
Dropout
StochasticDepth
Loss
MeanSquaredError
SoftmaxCrossEntropy
Datasets
CIFAR-10
CIFAR-100
MNIST
Street View House Numbers (SVHN)
Summary
References
5. Model Testing and Training
Accuracy
Timing
Common stacks
Summary
6. Loading and Saving Models
Loading models
Saving models
Model size
Summary
7. Sample Deep Learning Tests
A simple XOR problem
Complete source code
Output
A penny for your thoughts
A simple XOR problem (part 2)
Complete source code
Output
Recurrent Neural Network Language Models (RNNLM)
Complete source code
Vocabulary
Output
Word prediction test
Complete source code
Output
Decoupled Neural Interfaces using Synthetic Gradients
Output
MNIST accuracy tester
Complete sourc