structured_inhibition_spatial_network_model/scripts/spatial_packing_entropy/analyse_spatial_layouts.py

import h5py
import numpy
import pandas

import scripts
from scripts.interneuron_placement import plot_neural_sheet, Pickle
from scripts.spatial_network.run_entropy_maximisation_orientation_map import get_orientation_map
from scripts.spatial_packing_entropy.sample_spatial_layout import data_folder, correlation_length, map_seed, \
    sheet_size, NE, NI, long_axis, short_axis, number_of_bins, ds

if __name__ == "__main__":
    with h5py.File(data_folder + "spatial_layouts.hdf5", "r") as f:
        sampling_result_dict = {
            "seeds": f["sampling"]["seeds"][:],
            "mean_entropies": f["sampling"]["entropies"][:],
            "overlaps": f["sampling"]["overlaps"][:]
        }

        generation_indices = sorted([int(key) for key in f["differential_evolution"]["1"].keys()])
        evolution_result_dict = {
            "generation": generation_indices,
            "mean_entropies": numpy.array([f["differential_evolution"]["1"][str(generation_idx)]["entropy"][()] for \
                generation_idx
                               in
                               generation_indices]),
            "overlaps": numpy.array([f["differential_evolution"]["1"][str(generation_idx)]["overlap"][()] for \
                generation_idx in
                         generation_indices])
        }

        max_entropy_result_dict = {
            "mean_entropy": f["max_entropy"]["entropy"][()] ,
            "overlap": f["max_entropy"]["overlap"][()],
            "orientations": f["max_entropy"]["orientations"][:]
        }


        best_orientation = f["differential_evolution"]["1"][str(generation_indices[-1])]["orientations"][:]

    orientation_map = get_orientation_map(correlation_length, map_seed, sheet_size, NE, data_folder=data_folder)
    spatial_layout = scripts.spatial_layout.SpatialLayout(orientation_map, NE, NI, long_axis, short_axis, sheet_size)
    circular_radius = numpy.sqrt(spatial_layout.long_axis * spatial_layout.short_axis)
    circular_layout = scripts.spatial_layout.SpatialLayout(orientation_map, NE, NI, circular_radius, circular_radius,
                                                           sheet_size)

    df = pandas.DataFrame(sampling_result_dict)
    import matplotlib.pyplot as plt

    fig, ax = plt.subplots(1, 1, figsize=(3, 4))
    ax_overview = ax
    ax_overview.scatter(df["overlaps"].values/1000.0, df["mean_entropies"].values, marker='.', label='rand')
    ax_overview.plot(evolution_result_dict["overlaps"]/1000.0, evolution_result_dict["mean_entropies"], 'r-o',
                     label="evolv")
    ax_overview.plot(max_entropy_result_dict["overlap"]/1000.0, max_entropy_result_dict["mean_entropy"], 'r^',
                     label="max ent")
    ax_overview.plot(circular_layout.get_overlap(numpy.zeros(NI,), ds)/1000.0, circular_layout.get_mean_entropy(
        numpy.zeros(
        (NI,)), number_of_bins), 'rv', label="circ")
    ax_overview.set_xlabel("overlap (um2) x0.001")
    ax_overview.set_ylabel("<entropy>")
    ax_overview.legend()
    fig.tight_layout()
    fig.savefig("../../figures/evolve_spatial_layout/overlap_vs_entropy.png", dpi=300, bbox="tight")



    plot_neural_sheet(spatial_layout.ex_positions, spatial_layout.ex_tunings, spatial_layout.get_axonal_clouds(best_orientation))
    plt.savefig("../../figures/evolve_spatial_layout/layout_max_entropy.png", dpi=300)

    plot_neural_sheet(spatial_layout.ex_positions, spatial_layout.ex_tunings,
                      spatial_layout.get_axonal_clouds(max_entropy_result_dict["orientations"]))
    plt.savefig("../../figures/evolve_spatial_layout/layout_overlap_constrained_max_entropy.png", dpi=300)

    plot_neural_sheet(circular_layout.ex_positions, circular_layout.ex_tunings,
                      circular_layout.get_axonal_clouds(numpy.zeros((NI,))))
    plt.savefig("../../figures/evolve_spatial_layout/layout_circular.png", dpi=300)