Commit 9a2c6b4d026288745fe95708e3ff55f00a2351fd
1 parent
890a775449
Exists in
master
New file that help generating some stats and distribution stats (via plots)
Showing 1 changed file with 114 additions and 0 deletions Inline Diff
volia/stats.py
| File was created | 1 | ||
| 2 | import argparse | ||
| 3 | |||
| 4 | import os | ||
| 5 | import core.data | ||
| 6 | import math | ||
| 7 | import numpy as np | ||
| 8 | import scipy.stats | ||
| 9 | import pickle | ||
| 10 | import matplotlib.pyplot as plt | ||
| 11 | import matplotlib.colors as mcolors | ||
| 12 | |||
| 13 | |||
| 14 | |||
| 15 | from cycler import cycler | ||
| 16 | |||
| 17 | def stats(): | ||
| 18 | print("Decisions") | ||
| 19 | |||
| 20 | |||
| 21 | print(list(mcolors.TABLEAU_COLORS)) | ||
| 22 | |||
| 23 | |||
| 24 | if __name__ == "__main__": | ||
| 25 | |||
| 26 | # Parser | ||
| 27 | parser = argparse.ArgumentParser(description="") | ||
| 28 | |||
| 29 | # Arguments | ||
| 30 | parser.add_argument("--predictions", type=str, help="prediction file", required=True) | ||
| 31 | parser.add_argument("--labels", type=str, help="label file", required=True) | ||
| 32 | parser.add_argument("--labelencoder", type=str, help="label encode pickle file", required=True) | ||
| 33 | parser.add_argument("--outdir", type=str, help="output file", required=True) | ||
| 34 | |||
| 35 | args = parser.parse_args() | ||
| 36 | |||
| 37 | predictions = core.data.read_id_values(args.predictions, float) | ||
| 38 | labels = core.data.read_labels(args.labels) | ||
| 39 | |||
| 40 | le = None | ||
| 41 | with open(args.labelencoder, "rb") as f: | ||
| 42 | le = pickle.load(f) | ||
| 43 | stats = {} | ||
| 44 | |||
| 45 | print("PREDICTIONS ---------------------------") | ||
| 46 | for id_, predictions_ in predictions.items(): | ||
| 47 | label = labels[id_][0] | ||
| 48 | if label not in stats: | ||
| 49 | stats[label] = { | ||
| 50 | "nb_utt": 1, | ||
| 51 | "predictions": np.expand_dims(predictions_, axis=0) | ||
| 52 | } | ||
| 53 | else: | ||
| 54 | stats[label]["nb_utt"] = stats[label]["nb_utt"] + 1 | ||
| 55 | stats[label]["predictions"] = np.append(stats[label]["predictions"], np.expand_dims(predictions_, axis=0), axis=0) | ||
| 56 | |||
| 57 | |||
| 58 | print("CALCULATING ---------------------------") | ||
| 59 | |||
| 60 | |||
| 61 | colors = [ | ||
| 62 | "darkorange", | ||
| 63 | "red", | ||
| 64 | "blue" | ||
| 65 | ] | ||
| 66 | custom_cycler = (cycler(color=list(mcolors.TABLEAU_COLORS)) * | ||
| 67 | cycler(linestyle=['-', '--', '-.'])) | ||
| 68 | |||
| 69 | |||
| 70 | for label, stats_ in stats.items(): | ||
| 71 | |||
| 72 | plt.gca().set_prop_cycle(custom_cycler) | ||
| 73 | stats_mean = np.mean(stats_["predictions"], axis=0) | ||
| 74 | stats_std = np.std(stats_["predictions"], axis=0) | ||
| 75 | |||
| 76 | #print(label) | ||
| 77 | #print(stats_mean) | ||
| 78 | #print(stats_std) | ||
| 79 | kwargs = dict(alpha=0.5) | ||
| 80 | |||
| 81 | for i in range(stats_["predictions"].shape[1]): | ||
| 82 | label_str = le.inverse_transform([i])[0] | ||
| 83 | #plt.hist(stats_["predictions"][:, i], bins=10, label=label_str, **kwargs) | ||
| 84 | mu = stats_mean[i] | ||
| 85 | variance = stats_std[i] * stats_std[i] | ||
| 86 | sigma = stats_std[i] | ||
| 87 | # math.sqrt(variance) | ||
| 88 | print(f"{i}: mu {mu}, var {variance}, sigma {sigma}") | ||
| 89 | |||
| 90 | #x_values = np.arange(-1, 5, 0.1) | ||
| 91 | |||
| 92 | #y_values = scipy.stats.norm(mu, variance) | ||
| 93 | #y = scipy.stats.norm.pdf(x,mean,std) | ||
| 94 | |||
| 95 | #plt.plot(x_values, y_values.pdf(x_values,)) | ||
| 96 | |||
| 97 | #x, step = np.linspace(mu - 3*sigma, mu + 3*sigma, 1000, retstep=True) | ||
| 98 | x = np.linspace(0, 1, 1000) | ||
| 99 | #x = np.linspace(mu - 3*sigma, mu + 3*sigma, 1000) | ||
| 100 | #x, step = np.linspace(0, 1, 1000, retstep=True) | ||
| 101 | |||
| 102 | P = scipy.stats.norm.cdf(x, mu, sigma) | ||
| 103 | #print(step) | ||
| 104 | plt.plot(x, P, label=label_str, **kwargs) | ||
| 105 | #plt.savefig("simple_gaussian.pdf") | ||
| 106 | |||
| 107 | plt.legend() | ||
| 108 | plt.savefig(os.path.join(args.outdir, f"{label}_prediction_cdf.pdf")) | ||
| 109 | plt.clf() | ||
| 110 | |||
| 111 | |||
| 112 | # TODO: | ||
| 113 | # One graph for each label. Distribution of their predictions output are displayed. | ||
| 114 | |||
| 115 |