A new parameter grid of PS-Tree

[hengzhe-zhang]

Hello! I have updated the parameter grid. In order to avoid prior bias, I use a random sampler to randomly sample some hyperparameters from the parameter grid. I don't know whether this design will perform better or not than my baseline hyperparameter because hyperparameter sampling is a rather stochastic process. Thus, I'm looking forward to the benchmark result.

[lacava]

the one issue I see is that the random_state of ParameterSampler is uncontrolled. Is there a way to avoid that?

[hengzhe-zhang]

@lacava I have updated my code to add a random seed for the random sampler.