Evolutionary Deep Learning: Genetic algorithms and neural networks by Micheal Lanham.
Today Michael is a Lead AI Developer at Brilliant Harvest Inc. and also a Technical author at Manning Publications. Previously he worked at Symend as a Principle AI Engineer and Manager of ML Engineering.
This books expands upon his knowledge in deep learning, evolutionary computation, and genetic algorithms. This was a very enlightening read and will be very interesting for those interested in a comprehensive understanding of deep learning today.
The book is certainly written for data scientists using Python. This offers a very deep perspective on the combination of evolutionary principles with deep learning.
This results in enhanced model performance with the ability to solve complex problems with machine learning.
Michael is able to address multiple themes related to the link with computation and deep learning techniques. The book is structured in three main parts:
- Getting Started:
Introduces evolutionary deep learning, evolutionary computation, and genetic algorithms using DEAP (Distributed Evolutionary Algorithms in Python). - Optimizing Deep Learning:
Covers automated hyperparameter optimization, neuroevolution optimization, and evolutionary convolutional neural networks. - Advanced Applications:
Explores evolving autoencoders, generative deep learning with evolution, NeuroEvolution of Augmenting Topologies (NEAT), and future directions in evolutionary machine learning..
Key themes to learn
Michael addresses the following key themes include principles based upon biological evolution:
- Evolutionary Computation (EC) and Deep Learning Integration: How do EC principles enhance deep learning models? Michael is offering a view that combines biology-inspired algorithms with neural networks.
- AutoML and Hyperparameter Optimization:
This in fact, addresses techniques for automating the optimization of deep learning hyperparameters by using evolutionary computation, genetic algorithms, and particle swarm optimization. - Neuroevolution:
So neuroevolution optimization and evolutionary convolutional neural networks, reveal how evolutionary principles can be applied to neural network architectures. - Generative Deep Learning:
Certainly the core intersection of generative deep learning and evolution. Michael shows how these concepts work together. - NEAT (NeuroEvolution of Augmenting Topologies):
A method for evolving neural network topologies along with weights. This provides to be very key to the success of LLMs. - Reinforcement Learning:
Basic reinforcement learning that includes Q-Learning, and how to apply these concepts to deep learning. - Unsupervised Learning:
Michael is certainly addressing one of the key elements of machine learning: the optimization of unsupervised autoencoders which does include loss functions and network architectures. - Practical Applications:
Finally, a series of hands-on examples which permits readers to experiment with these techniques. Not for the faint of heart however.
Consider these books which also are excellent reads on Deep Learning: The Deep Learning Revolution and Deep Learning
In conclusion, this book is absolutely fascinating. Please let me share that I am also find swarm optimization very interesting and certainly an subject area that I would like to further understand.