"Logging attribute weights with ProcessLog"
Guru
When using EvolutionaryWeighting or a similar operator, is there a way to log the attribute weights that are being tested in each iteration with ProcessLog?
I'm thinking of something analogous to how the "feature_names" value works for feature selection processes. That way you could see what the weights that produced each performance value logged. E.g., if you were run a weighting process with three attributes, and logged generation, rms error, and the proposed "feature_weights", you'd see the following:
GEN RMS WGTS
1 10.5 0.10,0.25,0.30
1 11.2 0.11,0.22,0.39
2 10.2 0.09,0.30,0.29
2 13.4 0.15,0.20,0.30
Thanks,
Keith (... who's hoping he didn't overlook something trivial...)
I'm thinking of something analogous to how the "feature_names" value works for feature selection processes. That way you could see what the weights that produced each performance value logged. E.g., if you were run a weighting process with three attributes, and logged generation, rms error, and the proposed "feature_weights", you'd see the following:
GEN RMS WGTS
1 10.5 0.10,0.25,0.30
1 11.2 0.11,0.22,0.39
2 10.2 0.09,0.30,0.29
2 13.4 0.15,0.20,0.30
Thanks,
Keith (... who's hoping he didn't overlook something trivial...)
Tagged:
0
Answers
no, logging the feature weights is currently not easily possible, sorry.
Cheers,
Ingo