Info
This post was imported from a personal note. It may contain inside jokes, streams of consciousness, errors, and other nonsense.
The Eigen site is still down but I finally changed the code in the NeuralNetwork class to use Eigen matrix and vector classes instead of multiplying and adding floats by hand. I don’t see any performance improvement from running evaluate with 30 generations but I also haven’t bothered to benchmark or use the flame graph. Even if there’s no performance improvement, at least it opens the door to me doing bigger neural networks.
Now I can add a neuron indicating the boid’s current energy level and see if that results in different behaviour.
I tried it. I saw some of the boids straighten out a little more as they approached death rather than keep going in circles. *shrug*