3 ms·
I’m sure you could find a way to do GAs on a GPU. If you could have generations measured in billions of individuals and compute all of their fitness, mutation,
by throwaway316943 6y ago
I’m sure you could find a way to do GAs on a GPU. If you could have generations measured in billions of individuals and compute all of their fitness, mutation, crossover in parallel it should drastically speed up the algorithm.
- andretti1977 6y agoThat's exactly the point: is it possible? From my understandings, GPU allows massive parallel matricial calculations but that's not the case for mutations and crossovers (which involves random selection of mutate gene). Even so complex fitness functions (that are based on mathematical operation that takes into account more data per point like a moving average) couldn't be converted to matricial calculus.