HierarchicalDSR/plotter.py at main · DurstewitzLab/HierarchicalDSR · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
from matplotlib import pyplot as plt
import numpy as np
import torch


class Plotter():
    """Class for plotting the results of the model."""
    def __init__(self, model, dataset):
        """Initializes the plotter.
        Args:
            model: model to plot the results for
            args: args from the command line
            dataset: dataset that contains the test set
        """
        self.model = model
        self.dataset = dataset
        # set model to eval mode
        self.model.eval()

    @torch.compiler.disable()
    def plot_fixed_points(self):
        generated = self.model.evaluator.get_gen_data()
        N, T, dx = generated.shape
        n = int(np.ceil(np.sqrt(N)))
        if dx == 3:
            fps = self.model.evaluator.get_scyfi()
            fig = plt.figure(figsize=(3*n, 3*n), layout='compressed')
            for i, gen in enumerate(generated):
                gen = gen.cpu().numpy()
                ax = fig.add_subplot(n, n, i+1, projection='3d')
                ax.plot(gen[:,0], gen[:,1], gen[:,2])
                if fps[i] is not None:
                    ax.scatter(fps[i][:,0], fps[i][:,1], fps[i][:,2], marker='x', color='black')
                ax.set_title("Subject {}".format(i))
            return fig
        else:
            return None

    @torch.compiler.disable()
    def plot_power_spectrum(self):
        """Plots the power spectrum of the test set and the generated
        trajectories.
        """
        # compute power spectrum for first time, they are then saved within evaluator
        gen_ps, gt_ps = self.model.evaluator.get_power_spectrum()
        N, f, dz = gen_ps.shape
        interval = self.model.evaluator.get_interval_of_interest()
        fig = plt.figure(figsize=(4*N, self.model.dz), layout='compressed')
        sfigs = fig.subfigures(1, N)
        for i, (gt, gen) in enumerate(zip(gt_ps, gen_ps)):
            sfig = sfigs[i] if N > 1 else sfigs
            for j in range(gt.shape[-1]):
                ax = sfig.add_subplot(gt.shape[-1], 1, j+1)
                ax.plot(gt[:,j], label="ground truth")
                ax.plot(np.linspace(0, len(gt), len(gen)), len(gen)/len(gt) * gen[:,j], label="generated")
                ax.set_xlim(interval[i,j])
            sfig.suptitle(f"Subject {i}")
        return fig

    @torch.compiler.disable()
    def plot_trajectory(self):
        """Plots a test trajectory and a generated trajectory
        of same length.
        """
        ground_truth = self.dataset.get_test_data()
        N, T, dz = ground_truth.shape
        ground_truth = self.model.evaluator.get_test_data()[:, :T]
        generated = self.model.evaluator.get_gen_data()[:,:T]
        fig = plt.figure(figsize=(4*N, self.model.dz), layout='compressed')
        sfigs = fig.subfigures(1, N)
        for i, (gt, gen) in enumerate(zip(ground_truth, generated)):
            gt = gt.cpu().numpy()
            gen = gen.cpu().numpy()
            sfig = sfigs[i] if N > 1 else sfigs
            for j in range(gt.shape[-1]):
                ax = sfig.add_subplot(gt.shape[-1], 1, j+1)
                ax.plot(gt[:,j], label="ground truth")
                ax.plot(gen[:,j], label="generated")
                ax.legend()
            sfig.suptitle(f"Subject {i}")
        return fig

    @torch.compiler.disable()
    def plot_3D_trajectory(self):
        """Plots a test trajectory and a generated trajectory
        of same length in 3D.
        """
        if self.model.dx != 3:
            return None
        ground_truth = self.dataset.get_test_data()
        N, T, dz = ground_truth.shape
        n = int(np.ceil(np.sqrt(N)))
        generated = self.model.evaluator.get_gen_data()[:,:T]
        fig = plt.figure(figsize=(3*n, 3*n), layout='compressed')
        for i, (gt, gen) in enumerate(zip(ground_truth, generated)):
            gt = gt.cpu().numpy()
            gen = gen.cpu().numpy()
            ax = fig.add_subplot(n, n, i+1, projection='3d')
            ax.plot(gt[:,0], gt[:,1], gt[:,2], label="ground truth")
            ax.plot(gen[:,0], gen[:,1], gen[:,2], label="generated")
            ax.set_title("Subject {}".format(i))
            ax.legend()
        return fig

    @torch.compiler.disable()
    def plot_hovmoller(self):
        """Plots a test trajectory and a generated trajectory
        of same length in a hovmoller diagram."""
        ground_truth = self.dataset.get_test_data()
        N, T, dz = ground_truth.shape
        generated = self.model.evaluator.get_gen_data()[:,:T]
        fig = plt.figure(figsize=(9*N, int(T/100)), layout='compressed')
        for i, (gt, gen) in enumerate(zip(ground_truth, generated)):
            gt = gt.cpu().numpy()
            gen = gen.cpu().numpy()
            ax = fig.add_subplot(N, 1, i+1)
            # concat trajectories along latent dim
            trajectory = np.concatenate([gt, gen], axis=-1)
            c = ax.imshow(trajectory.T, origin='lower', cmap='bwr', aspect='auto')
            ax.axhline(dz, color='black', linewidth=1)
            fig.colorbar(c, ax=ax)
            ax.set_title("Subject {}".format(i))
            ax.set_ylabel('T')
        return fig

    @torch.compiler.disable()
    def plot_trajectory_train(self):
        """Copy of plot_trajectory but uses a single train instance
        as ground truth
        """
        ground_truth = self.dataset[0][0][None]
        N, T, dz = ground_truth.shape
        generated = self.model.evaluator.get_gen_data()[:,:T,:]
        fig = plt.figure(figsize=(3*N, self.model.dz))
        sfigs = fig.subfigures(1, N, layout='compressed')
        for i, (gt, gen) in enumerate(zip(ground_truth, generated)):
            gt = gt.cpu().numpy()
            gen = gen.cpu().numpy()
            sfig = sfigs[i] if N > 1 else sfigs
            for j in range(gt.shape[-1]):
                ax = sfig.add_subplot(gt.shape[-1], 1, j+1)
                ax.plot(gt[:,j], label="ground truth")
                ax.plot(gen[:,j], label="generated")
                ax.legend()
            sfig.suptitle(f"Subject {i}")
        return fig

    def plot_hierarchisation_stuff(self):
        """Plots interesting things from the hierarchisation scheme."""
        return self.model.hierarchisation_scheme.plot_stuff()