Website Source: blog / neat_jax_blog

Summary

Pending synthesis from local website source.

Original source title: Shoving NEAT into JAX: When Evolving Topologies Meet Static Computation Graphs

Extracted Preview

Note: This started as a passion project. It remains a passion project. I have made peace with this.

Shoving NEAT into JAX: When Evolving Topologies Meet Static Computation Graphs

Project Home: [github.com/yash-srivastava19/NEAT-JAX](https://github.com/yash-srivastava19/NEAT-JAX)

---

NEAT is from 2002. JAX is from 2018. Making them talk to each other taught me more about both than reading any paper did.

---

What is NEAT, and why should you care?

NEAT - Neuroevolution of Augmenting Topologies - is a genetic algorithm for evolving neural networks, introduced by Stanley and Miikkulainen in 2002. What makes it interesting, and what makes it hard, is that it doesn't just evolve *weights*. It evolves *topology*. The structure of the network itself - how many nodes, how they're connected - changes over generations.

The life cycle of a NEAT population looks like this:

Integration Notes

Source section: blog
Local source: /home/yashs/Desktop/Programming/yash_blog/yash-srivastava19.github.io/blog/neat_jax_blog.md
Raw copy: raw/website/yash-srivastava19-github-io/blog/neat_jax_blog.md