Hacker News Reader

I wonder how many people actually read this 40 page doc having seen this link. Would be nice if the OP gave us a clue why they posted it, perhaps a summary.

View on HN

I read a substantial fraction of it before submitting it.

I'm interested in evolutionary computation because it's simple to implement and historically has competed with neural networks.

They offer the potential to create interpretable models.

The method of optimization probably has little to do with the interpretability of the model. Gradient optimization (i.e. backpropagation) can be used to fit logistic regression models, which are interpretable. Evolutionary algorithms can be used to fit deep neural networks, and will not make it more interpretable.

I'd say the least intepretable systems are biological, which are the outcome of an evolutionary process

Reply on news.ycombinator.com