A first step towards using pandas

numbbo / coco

Numerical Black-Box Optimization Benchmarking Framework

https://numbbo.github.io/coco

Other

260 stars 86 forks source link

A first step towards using pandas #1927

Open nikohansen opened 4 years ago

nikohansen commented 4 years ago

We want a module that reads in (or converts) data from one (or several) full experiment(s) into a pandas DataFrame (or possibly several data frames).

This may be useful in particular, because from there it could be much easier to create "from scratch" some (new) desired output functionality.

nikohansen commented 3 years ago

The reverse may also be valuable, as suggested by @MLopez-Ibanez : allow to feed a "standard" pandas DataFrame into a DataSetList from which all output is generated.

nikohansen commented 4 months ago

The first step going bottom-up may be:

[ ] implement a DataSet.to_frame_data method that returns a list of n_instances data each of them can serve as argument to the pandas.DataFrame like

    frames = {}
    for instance, data in data_set.to_frame_data().items():
        frames[instance] = pd.DataFrame(data, columns=['evaluations', 'targets'])