machine learning - When encoding weights in a neural network as a chromosome in a genetic algorithm, can a binary string be too long to function properly? -


i have feedforward neural network want train using genetic algorithm. have read best option use binary string of weights represented grey codes. in case, 65 weights each chromosome, result in string of length 2080 (65*32 bits). understand complex problem, take longer reach optimal solution having smaller number of bits in string, 2080 long ga work @ all? there better way encode such large number of weights?

i don't think size of string of problem, may problem-dependent.

if worried size of strings, perhaps reduce precision lower number of bits per weight , observe effects has on learning performance. have stated, grey codes best representation of weights. i've used ga's in other application areas gene sizes around same length , have evolved well.

of course, need ensure population size , number of generations sufficient enough problem , fitness function.


Comments

Popular posts from this blog

php - Submit Form Data without Reloading page -

linux - Rails running on virtual machine in Windows -

php - $params->set Array between square bracket -