Hyppopy
Hyppopy copied to clipboard
Number Shift in the resulting Dataframe
I specified the following parameter Space and ran the Optunity Solver: config = { "hyperparameter": { "convlayers": { "domain": "uniform", "data": [1, 5], "type": int }, "denselayers": { "domain": "uniform", "data": [1, 3], "type": int } } For every run recorded in the Pandas Dataframe the Index of the number is displayed, not the actual number. For example convlayers: 0, but should be convlayers: 1 and so on. I think it is preferable to store the actual integer number
Hi yannick,
thanks for the input. As far as I see it now, I would deem it a bug, as it breaks my expectations, because I would also expect the use hyper parameter value (and not propably its index on the axis). But we have to have a closer look.
Could you do me a favor and test a nother optimizer with your setting. That would help to narrow it down (either a problem in the optunity wrapper or the hyppopy "managment" layer). Thanks.
I have a similar issue when using the Hyperopt Solver. The issue is only occuring for my hyperparameter of type 'int'. Type 'float' does not seem to be affected by this bug in my case.