`seed=0` might not fix the optimisation result
nzw0301 opened this issue · comments
Kento Nozawa commented
When I use seed=0
to fix the optimisation results, it works except for seed=0
. However, when I used seed=0
, the results differ. The minimal reproducible code would be as follows:
import cma
es = cma.CMAEvolutionStrategy(8 * [0], 0.5, inopts={"seed": 0})
es.optimize(cma.ff.rosen)
I suppose the first condition at
pycma/cma/evolution_strategy.py
Line 1487 in 5e45183
False
when seed=0
, as a result, the seed value is not used as I expected.
Related versions:
Python: 3.10.9
pycma: 3.3.0
Kento Nozawa commented
Sorry, this was the intended behaviour according to #111; let me close this issue.