Frugal learning in neural networks

Project description

The brain is one of the most energy intense organs. Some of this energy is used for neural information processing, however, fruitfly experiments have shown that also learning is metabolically costly. We have developed a number of ways for neural networks to learn while spending less energy on learning than standard training methods like back prop.
See recent papers on https://scholar.google.com/citations?user=Juxhhu0AAAAJ

In this project we continue this search for frugal learning algorithms. In particular we aim to derive minimum bounds on the energy need to change a network's function from one to another.

The project will be a mix of Python coding and analytical work. 
Background in biology is not required.