Feat/quick density plot

[quentinmarolleau]

• 104

• data treatment performed by pandas package
• visualisation performed by seaborn package
• data sharing the same parameters are "stacked" on a same "cell" → allows the computation of average and std values for each set of parameters
• the heatmaps of average values, and standard dev (with size of the sample in parenthesis) are plotted side-by-side, or on top of each other (depending of the aspect ratio of the heatmaps)

e.g. with fake data:

[adareau]

OK, looks nice. I'm just curious: do you really have to use pandas et seaborn, rather than numpy/matplotlib ? The point is just to avoid having multiple packages & standards across HAL to handle data & plots...

[quentinmarolleau]

OK, looks nice. I'm just curious: do you really have to use pandas et seaborn, rather than numpy/matplotlib ? The point is just to avoid having multiple packages & standards across HAL to handle data & plots...

probably not, but it is MUCH simpler to manipulate dataframes to do this kind of stuff (especially with the groupby method of pd.DataFrames). Btw pip install seaborn will also install pandas.

Additionaly, the plot is ultimately handled with matplotlib since seaborn is just a kind of add-on that makes the visualisations of pandas dataframes easier, and is then piped into a matplotlib figure/axes.