add license and plots

Author: IraKorshunova

LICENSE

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2014 Iryna Korshunova
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

plots/__init__.py

@@ -0,0 +1 @@
+__author__ = 'user'

plots/plot_examples.py

@@ -0,0 +1,52 @@
+import numpy as np
+import json
+import os
+import cPickle
+import copy
+
+import matplotlib.pyplot as plt
+from theano import config
+
+from utils.loader import load_train_data
+from utils.config_name_creator import *
+from utils.data_scaler import scale_across_features, scale_across_time
+
+
+config.floatX = 'float32'
+
+
+def plot_examples(subject, data_path, scale):
+    d = load_train_data(data_path, subject)
+    x, y, filename_to_idx = d['x'], d['y'], d['filename_to_idx']
+    if scale:
+        x, _ = scale_across_time(x=x)
+
+    for filename, idx in filename_to_idx.items():
+        fig = plt.figure()
+        fig.suptitle(filename)
+        for i in range(x[idx].shape[0]):
+            fig.add_subplot(4, 4, i)
+            plt.imshow(x[idx, i, :, :], aspect='auto', origin='lower', interpolation='none')
+            plt.colorbar()
+        plt.show()
+
+
+if __name__ == '__main__':
+
+    with open('SETTINGS.json') as f:
+        settings_dict = json.load(f)
+
+    # path
+    data_path = settings_dict['path']['processed_data_path'] + '/' + create_fft_data_name(settings_dict)
+    write_dir = data_path + '/img'
+    if not os.path.exists(write_dir):
+        os.mkdir(write_dir)
+
+    # params
+    model_params = settings_dict['model']
+    validation_params = settings_dict['validation']
+
+    names = ['Dog_1', 'Dog_3', 'Dog_2', 'Dog_5', 'Dog_4', 'Patient_1', 'Patient_2']
+    for subject in names:
+        print '***********************', subject, '***************************'
+        plot_examples(subject, data_path, scale=False)

plots/plot_filters.py

@@ -0,0 +1,39 @@
+import json
+from utils.config_name_creator import create_cnn_model_name
+import numpy as np
+import matplotlib.pyplot as plt
+import cPickle
+import PIL.Image as Image
+
+
+def paint_filter(patient_name, model_path):
+    with open(model_path + '/' + patient_name + '.pickle', 'rb') as f:
+        state = cPickle.load(f)
+
+    # first layer
+    W1 = state['weights'][0]
+    width, heights = W1.shape[3], W1.shape[2]
+    n_filters = W1.shape[0]
+    n_fbins = state['params']['n_fbins']
+    print W1.shape
+    x = np.zeros((heights, width * n_filters))
+    for i in range(n_filters):
+        x[:, i] = np.reshape(W1[i, 0, :, :], heights * width)
+
+    ax = plt.gca()
+    ax.set_yticks(range(0, heights, n_fbins))
+    ax.yaxis.grid(True, which='major', linestyle='-')
+    plt.imshow(x, interpolation='none')
+    plt.show()
+
+
+if __name__ == '__main__':
+    with open('SETTINGS.json') as f:
+        settings_dict = json.load(f)
+
+    s1 = '0.81448_nfreq6featumeanloghighc180lowcu0.1win_l60strid60globa0recep[1, 2]use_t0activ[urelu, urelu, urelu]dropo[0.2, 0.5]overl0strid[1, 1]train10weigh0.01scale0nkern[16, 32, 128]pool_[1, 1]l2_re0.0001valid10max_i150000rando1'
+    s2 = '0.80192_nfreq8featumeanlog_stdhighc180lowcu0.1win_l120strid120globa1recep[1, 1]use_t0activ[urelu, urelu, utanh]dropo[0.3, 0.6]overl4strid[1, 1]train10weigh0.01scale1nkern[16, 32, 512]pool_[1, 1]l2_re0.0001valid10max_i150000rando1'
+    model_path = settings_dict['path']['model_path'] + '/' + s2  # create_cnn_model_name(settings_dict)
+    names = ['Dog_1', 'Dog_2', 'Dog_3', 'Dog_4', 'Dog_5', 'Patient_1', 'Patient_2']
+    for patient_name in names:
+        paint_filter(patient_name, model_path)

plots/plot_hour_means.py

@@ -0,0 +1,68 @@
+import numpy as np
+import json
+import os
+import cPickle
+import copy
+
+import matplotlib.pyplot as plt
+from theano import config
+
+from utils.loader import load_train_data
+from utils.config_name_creator import *
+from utils.data_scaler import scale_across_features, scale_across_time
+
+
+config.floatX = 'float32'
+
+
+def plot_hour_means(x, filename_to_idx, hours, type, write_dir):
+    for n_hour, hour in enumerate(hours):
+        mean_fft = np.zeros((x.shape[1], x.shape[2], x.shape[3]))
+        for clip in hour:
+            idx = filename_to_idx[clip]
+            mean_fft += x[idx]
+        mean_fft /= len(hour)
+
+        fig = plt.figure()
+        fig.suptitle(type+str(n_hour))
+        for i in range(mean_fft.shape[0]):
+            fig.add_subplot(4, 4, i)
+            plt.imshow(mean_fft[i, :, :], aspect='auto', origin='lower', interpolation='none')
+            plt.colorbar()
+        plt.show()
+        #plt.savefig(write_dir + '/' + subject + type + str(n_hour) + '.jpg')
+
+
+def plot(subject, data_path, write_dir, scale):
+    d = load_train_data(data_path, subject)
+    x, y, filename_to_idx = d['x'], d['y'], d['filename_to_idx']
+    if scale:
+        x, _ = scale_across_time(x=x)
+
+    filenames_grouped_by_hour = cPickle.load(open('filenames.pickle'))
+    preictal_hours = copy.deepcopy(filenames_grouped_by_hour[subject]['preictal'])
+    interictal_hours = copy.deepcopy(filenames_grouped_by_hour[subject]['interictal'])
+
+    plot_hour_means(x, filename_to_idx, preictal_hours, type='preictal', write_dir=write_dir)
+    plot_hour_means(x, filename_to_idx, interictal_hours[:10], type='interictal', write_dir=write_dir)
+
+
+if __name__ == '__main__':
+
+    with open('SETTINGS.json') as f:
+        settings_dict = json.load(f)
+
+    # path
+    data_path = settings_dict['path']['processed_data_path'] + '/' + create_fft_data_name(settings_dict)
+    write_dir = data_path + '/img'
+    if not os.path.exists(write_dir):
+        os.mkdir(write_dir)
+
+    # params
+    model_params = settings_dict['model']
+    validation_params = settings_dict['validation']
+
+    names = ['Dog_1', 'Dog_3', 'Dog_2', 'Dog_5', 'Dog_4', 'Patient_1', 'Patient_2']
+    for subject in names:
+        print '***********************', subject, '***************************'
+        plot(subject, data_path, write_dir, scale=True)

plots/plot_train_test.py

@@ -0,0 +1,59 @@
+import numpy as np
+import json
+from theano import config
+import cPickle
+from utils.loader import load_grouped_train_data, load_train_data, load_test_data
+from utils.config_name_creator import *
+from matplotlib import cm
+import matplotlib.pyplot as plt
+
+config.floatX = 'float32'
+
+
+def plot_train_test_sizes(subjects, data_path):
+    filenames_grouped_by_hour = cPickle.load(open('filenames.pickle'))
+    dict_train, dict_test = {}, {}
+    for subject in subjects:
+        x_train = load_train_data(data_path, subject)['x']
+        x_test = load_test_data(data_path, subject)['x']
+        dict_train[subject] = x_train.shape[0]
+        dict_test[subject] = x_test.shape[0]
+
+    plt_data = [dict_train, dict_test]
+    data_orders = [subjects, subjects]
+    colors = [cm.Greys(1. * i / (len(subjects) + 2)) for i in range(len(subjects))]
+
+    values = np.array([[data[name] for name in subjects] for data, order in zip(plt_data, data_orders)])
+    lefts = np.insert(np.cumsum(values, axis=1), 0, 0, axis=1)[:, :-1]
+    orders = np.array(data_orders)
+    bottoms = np.array([0.1, 0.8])
+
+    for name, color in zip(subjects, colors):
+        idx = np.where(orders == name)
+        value = values[idx]
+        left = lefts[idx]
+        plt.bar(left=left, height=0.3, width=value, bottom=bottoms,
+                color=color, orientation="horizontal", label=name)
+        left_margin0 = left[0] if value[0] < 100 else left[0] + value[0] / 3.0
+        left_margin1 = left[1] if value[1] < 100 else left[1] + value[1] / 3.0
+        plt.text(left_margin0, bottoms[0] + 0.15, value[0])
+        plt.text(left_margin1, bottoms[1] + 0.15, value[1])
+        if value[0] < 100:
+             ratio = str(len(filenames_grouped_by_hour[name]['interictal'])) + '/\n' + str(len(filenames_grouped_by_hour[name]['preictal']))
+        else:
+            ratio = str(len(filenames_grouped_by_hour[name]['interictal'])) + '/' + str(len(filenames_grouped_by_hour[name]['preictal']))
+        plt.text(left_margin0, bottoms[0]+0.08, ratio)
+
+    plt.yticks(bottoms + 0.15, ["train", "test"])
+    plt.legend(loc="best", bbox_to_anchor=(1.0, 1.00))
+    plt.subplots_adjust(right=0.85)
+    plt.show()
+
+
+if __name__ == '__main__':
+    with open('SETTINGS.json') as f:
+        settings_dict = json.load(f)
+
+    data_path = settings_dict['path']['processed_data_path'] + '/' + create_fft_data_name(settings_dict)
+    subjects = ['Dog_1', 'Dog_2', 'Dog_3', 'Dog_4', 'Dog_5', 'Patient_1', 'Patient_2']
+    plot_train_test_sizes(subjects, data_path)