CS 506 - Computational Geometry
This project will show the feasibility of using multiplicative weights update to explore for optima in regions of multiple fitness peaks, which are characterized by directional asymmetry induced by epistatic interaction between mutations in a genome.
We utilize data from a study of E. coli in order to ground-truth the epistatic interactions of beneficial mutations.
We model epistasis as a multiplicative pairwise interaction.
We compare this to simulated epistasis acting on an epistatic dataset which has positional interaction (as well as additive weights).
Finally, we compare this with the observed weight matrix perturbations observed in optimizing a system of mutated program edits (which we believe have positional epistasis).