Integrated tuning distributions BVC class

[jquinnlee]

Integrated tuning distribution options to determine preferred wall distances from params argument "pref_wall_dist_distribution". Allows user to select from uniform (new default), Rayleigh (previous default), normal, and delta distributions that will be used to generate wall distances for all BVCs in this class.

[jquinnlee]

Hey Tom,

Before I push any changes can you clarify the concern further? If I were to use: np.uniform(low=0, high=-.15, size=10) this will return 10 values uniformly distributed (flat) between 0 and -.15.

To get around the possible issue withself.Agent.Environmnet.scale in those environments with weird aspect ratios and allowing negative values for self.pref_wall_dist(which I think your wanting to allow?), what would you think of this:

if self.pref_wall_dist_distribution == 'uniform':
            dists = np.array([self.Agent.Environmnet.scale, self.pref_wall_dist * 2])
            self.tuning_distances = np.random.uniform(low=0, high=dists[np.argmin(np.abs(dists))], size=self.n)

Let me know if I am understanding your concerns or not, happy to make whatever changes meet your goals! The main concern I have with the function you suggested is that it will return a truncated normal distribution, and not a uniform distribution (which is the option I suggest adding).

[TomGeorge1234]

Oh sorry!! I mean the "normal" not the "uniform" option. The "normal" function could return a negative wall preference which mustn't be allowed.

So the uniform option should just be as you wrote it but with wall_pref_dist instead of self.Agent.Environment.scale:

if self.pref_wall_dist_distribution == 'uniform':
            self.tuning_dists = np.random.uniform(low=0,high=wall_pref_dist, n=self.n)
if self.pref_wall_dist_distribution == 'normal':
          # the code I wrote in my first response using scipy.truncnorm 

My bad, apologies for the confusion

[TomGeorge1234]

closing, instead merged #47