I gave a short talk at Microsoft Research Labs, titled “Introduction to Roach Infestation Optimization”. The goal of roach optimization is to find the values for constants in a prediction equation.
The standard technique is based on traditional calculus and is called gradient descent. Gradient descent has some technical problems so machine learning experts explore so-called bio-inspired algorithms.
Roach optimization loosely models some of the behaviors of roaches. In my talk I explained exactly how roach optimization works and gave listeners a few comment about other swarm intelligence algorithms.
One of the nicest things about working at Microsoft Research is that employees are actively encouraged to explore new ideas. This is currently called “having a growth mindset” but that’s just a label for an idea that’s been around for a long time and used to be called “continuous learning”.