Wrong scaling on hyperopt cross parameter plot with non-logarithmic parameters

[PAUL-BERNARD]

Cross-scatter plots rendered with plot_hyperopt_report only uses linear or logarithmic scale on both axis.

Here is a short reproducible code snippet that illustrates what's wrong

import reservoirpy
import random
import json
import math

reservoirpy.verbosity(0)

hyperopt_config = {
    "exp": "demo",
    "hp_max_evals": 600,
    "hp_method": "random",
    "instances_per_trial": 1,
    "hp_space": {
        "logarithmic_param": ["loguniform", 1e-5, 1e5],
        "linear_param": ["uniform", 0, 10]
    }
}

with open("./demo.json", "w+") as f:
    json.dump(hyperopt_config, f)

def objective(data, config, *, logarithmic_param, linear_param):
    return {'loss' : math.exp(random.uniform(-3, 0))}

reservoirpy.hyper.research(
    objective=objective,
    dataset=None,
    config_path="./demo.json",
    report_path="./"
)

reservoirpy.hyper.plot_hyperopt_report("./demo/", ("logarithmic_param", "linear_param"), not_log=['linear_param'])