search

orionw / configtune

An easy way to tune machine learning hyperparameters (especially for those that use a config file)
MIT License
1 stars 0 forks source link

Population Decay #6

Closed bjschoenfeld closed 5 years ago

bjschoenfeld commented 5 years ago

It seems that the population can sometimes dwindle from one generation to the next. In the tests I get:

INFO:tuningdeap.main:The population has a size of 8 after rejecting the invalid: should be 10

What is causing this? It does not seem like desired (default) behavior.

orionw commented 5 years ago

That info message should probably be renamed. It was there when I was creating the enforce_limits function but can be disregarded because if the population is less than population_size it'll regenerate more to replace the invalid individuals.

However, the population dwindling does seem to be a thing (outside of this info message) and should probably be looked into.

orionw commented 5 years ago

TODO: add a test to verify population size.