PCA_Weight.py

import os
import torch
import numpy as np
from matplotlib.cm import get_cmap
import matplotlib.pyplot as plt
from sklearn.decomposition import PCA, IncrementalPCA
from sklearn.externals import joblib

cmap = get_cmap('hsv')
plt.figure(figsize=(16, 7))

# All layers
all_params = []
all_accs = []
for i in range(1, 9):
        checkpoint_base = os.path.join(base, str(i))
        checkpoint_filenames = sorted(os.listdir(checkpoint_base))
        for epoch in range(2, epoch_max, epoch_step):
            filename = checkpoint_filenames[epoch]
            checkpoint = torch.load(os.path.join(checkpoint_base, filename))
            model = eval(checkpoint['arch'])()
            model.load_state_dict(checkpoint['state_dict'])
            params = np.zeros((0, ))
            for p in model.parameters():
                params = np.append(params, p.cpu().data.numpy().flatten())
            all_params.append(params)
            all_accs.append(checkpoint['logger'].entries[epoch-1]['accuracy'])

all_params = np.array(all_params)
all_params=IncrementalPCA(2, batch_size=6).fit_transform(all_params)
x = [p[0] for p in all_params]
y = [p[1] for p in all_params]

seg = len(range(1, epoch_max, epoch_step))
plt.subplot(121)
plt.title('All layers (DeepMnistCNN)')
for i in range(1, 9):
        x_ = x[(i - 1) * seg:i * seg]
        y_ = y[(i - 1) * seg:i * seg]
        a_ = all_accs[(i - 1) * seg:i * seg]
        plt.plot(x_, y_, 'o:', color=cmap((i - 1) / 7))
        for xi, yi, ai in zip(x_, y_, a_):
            plt.annotate(str('{:.1f}'.format(ai * 100)), xy=(xi, yi),
                         xytext=(xi + 0.004, yi + 0.004), color=cmap((i - 1) / 7.5))
plt.grid()