February 14 · 6:00 PM
Mostly all neural networks in machine learning today are stateless, always giving the same output 'y' for an input 'x'. Biological neurons on the other hand show more dynamic integration of input, with their output being similar to a binary on/off switch throughout time.
In this presentation you will learn about the similarities and distinctions between biological and artificial neurons, evolutionary algorithms, and fitness functions. You will also see a live simulation of these methods combined, where creatures evolve to find food in a randomized environment only using a simple spiking neural network to control them. As we will see, the end result of the creature's behavior is strikingly similar to that of small biological animals.
Presenter: Asgeir Berland from Inmeta
There will be beer.