Cover
Title Page
Copyright and Credits
About Packt
Contributors
Table of Contents
Preface
Section 1: Introduction
Chapter 1: Deep Learning for Finance 101
How AI is redefining the financial services industry
A brief history of AI in finance
A shared background (before 1880)
Computing probabilities (1880-1950)
Automatic reasoning Β (1950-1980)
Expert systems (1980-1990)
Narrow AI systems (1990-2010)
Machine learning at scale (2011-today)
A first look at deep learning for finance
Data gathering
Implementing an autoencoder
Using TensorFlow to implement the autoencoder
Summary
Chapter 2: Designing Neural Network Architectures
Going through the basics
Organizing neuronsΒ
Representational spaces
Learning the weights
Regularization
An interlude of history
Working with MLP
Neurons based on distance
Computing with tensors
Training a network through backpropagation
Understanding CNNs
LeNet-5, AlexNet, and others
Understanding RNNs
Long Short-Term Memory (LSTM)
Gated recurrent unit
Summary
Chapter 3: Constructing, Testing, and Validating Models
Building blocks of financial models
Formulating a hypothesis for trading
Selecting relevant financial models
Example β factor models for return prediction
Adding nonlinearity to linear models
Simple neural networks to capture non-linearity and preference shifts
DeepNets to incorporate memory in the modeling process
Machine learning versus statistical models
Acquiring data from multiple sources
Asynchronous
Revised or backfilled
Prone to manipulation
Outliers
Implementing the model
Keras
TensorFlow
Theano
Microsoft CNTK
PyTorch
Caffe2
MXNet
Chainer
Torch
Caffe
Wrappers
Evaluating investment strategy
Commonly used statistics
Commonly used financial metrics
Cumulative and monthly returns
Information coefficient
The information ratio and Sharpe ratio
Maximum drawdown
Sortino ratio
Tuning the model
Grid search
Random search
Bayesian optimization
Going live
Documenting investment strategy and code
Transitioning to a production environment
Paper portfolios
Soft launch
Go live!
Benchmarking
Benchmarking live data
Benchmarking to model diagnostics
Summary
Section 2: Foundational Architectures
Chapter 4: Index Replication by Autoencoders
Replicating an index
Data gathering
Implementing a vanilla AE
Data exploration and preparation
Creating and fitting the model
Evaluating the model
Replicating an index by using an AE
Exploring some AE variants
The denoising AE
The sparse AE
Understanding deep AE
Summary
Chapter 5: Volatility Forecasting by LSTM
Measuring volatility
Types of volatility
Historical volatility
Implied volatility
Volatility index
Intraday volatility
Realized volatility
Loading the data
Implementing the LSTM model
Data preparation
Creating and fitting the model
Evaluating the model
Improving the model's performance
Online learning
Stacking layers
Tuning the hyperparameters
Visualizing results
Comparing LSTM with other models
RNN model
The GARCH model
Visualizing the cumulative squared error
Summary
Chapter 6: Trading Rule Identification by CNN
Trading signals with technical indicators
Data handling
Getting data from public sources
Setting up the data
Hypothesis formulation and in-sample testing
Benchmarking alternative models
Benchmark 1 β simple trading rule
Benchmark 2Β β simple classification network
Constructing a convolutional neural network
Modeling investment logic
Selecting the network architecture
Setting up the data in the correct format
Training and testing the model
Summary
Section 3: Hybrid Models
Chapter 7: Asset Allocation by LSTM over a CNN
Modeling tactical asset allocationΒ
Defining our problem
Joint forecasting for an asset class
Individual forecasting and bets
Setting up data
Building a model
Understanding the deep learning model
Implementing a CNN-LSTM model
Testing and validating our model
Analyzing country models
Summary
Chapter 8: Digesting News Using NLP with BLSTM
Sentiment analysis for finance
Representing text data β words to vectors
Frequency-based word vectors
Count vectorization
TF-IDF vectorization
Word embeddings
Word2Vec
CBOW
Skip-gram
FastText
GloVe
Data loading and splitting
Implementing the BLSTM model
Data preparation
Creating and fitting the model
Evaluating the model
Improving performance
Dealing with imbalanced classes
Applying pre-trained word embeddings
Considering separate decisions
Summary
Chapter 9: Risk Measurement Using GAN
Estimating value at risk
Computing methods and drawbacks
Introducing generative adversarial networks
Generative models
Discriminative models
Inner workings of GAN
Implementing a risk model using GAN
Defining our model
Implementing the GAN model
Benchmarking results
Summary
Section 4: Advanced Techniques
Chapter 10: Chart Visual Analysis by Transfer Learning
Explaining what transfer learning is
Understanding transfer learning
What to transfer?
When to transfer?
How to transfer?
Using visual inspection in transfer learning for technical analysisΒ
What to transfer?
When to transfer?
How to transfer?
Implementing a transfer learning model
Acquiring and formatting data
Setting up data for the ResNet50 model
Importing and training the model
Predicting test images
Summary
Chapter 11: Better Chart Analysis Using CapsNets
Understanding CapsNets
Modeling CapsNets
Dynamic routing between capsules
Matrix capsules with EM routing
Advantages of CapsNets
Disadvantages of CapsNets
Constructing a CapsNet model
Implementing the model
Setting up data
Training the model
Summary
Chapter 12: Training Trader Robots Using Deep Reinforcement Learning
Understanding Reinforcement Learning
Deep Q-learning
Formulating the RL problem
State
Action
Reward
Configuring the data
Loading the data
Defining a trading strategy
Input data
Data preparation
Implementing a Robot based onΒ Deep Q-learning
Designing the agent
DQN
Remember
Experience replay
Act
Training the agent
Evaluating the model
Summary
Further Research
Chapter 13: What Next?
Automating discovering and learning models from dataΒ
Distributing computations across multiple computers and GPUs
Distributed deep learning
Data parallelism
Model parallelism
Layer pipelining
Frameworks for deep learning
Horovod
Distributed TensorFlow models
BigDL
Elephas
Exploiting deep learning for high-frequency trading
Using deep learning in other FinTech applications
Payment transfer and processing
Robo advisory
Alternate currencies
Concerns about risks and the future of deep learning in finance
Concluding remarks
Other Books You May Enjoy
Index