def main(): args = arg.preprocess_args() rand = random.randint(0, 9999999) data, _ = helper.pre_processed_data(args, rand, dry=False) # data = data.reshape(-1,48,48) if args.mean: label, _ = helper.pre_processed_label(args, rand, dry=False) data = preprocess.mean_image(label, data) elif args.split is not None or args.randomize: label, _ = helper.pre_processed_label(args, rand, dry=False) helper.write_data_to_file( (args.folder or helper.FOLDER) + (args.name or 'processed_data') + '_l' + helper.EXT, label, fmt='%d', h='0') # for i in range(4): # for j in range(4): helper.write_data_to_file( (args.folder or helper.FOLDER) + (args.name or 'processed_data') # + str(i) + str(j) + helper.EXT, data, # [k[i*12:(i*12)+12,j*12:(j*12)+12].flatten() for k in data], h=', '.join(str(i) for i in range(len(data[0]))))
def main(): args = arg.preprocess_args() rand = 0 data, _ = helper.pre_processed_data(args, rand, dry=False) if args.mean: rand = random.randint(0, 9999999) label, _ = helper.pre_processed_label(args, rand, dry=False) data = preprocess.mean_image(label, data) helper.create_images_from_rows((args.name or 'img'), data)
def main(): print('start clustering') args = kmeans_args() rand = np.random.randint(10000000) data_train, data_test = pre_processed_data(args, rand) label_train, label_test = pre_processed_label(args, rand) print('data loaded') kmeans(data_train, data_test, label_train, label_test, args) minibatchkmeans(data_train, data_test, label_train, label_test, args) featureagglomeration(data_train, data_test, label_train, label_test, args) birch(data_train, data_test, label_train, label_test, args) spectralclustering(data_train, data_test, label_train, label_test, args) print('done')
def main(): args = kmeans_args() rand = np.random.randint(10000000) data_train, data_test = pre_processed_data(args, rand) label_train, label_test = pre_processed_label(args, rand) print('data loaded') switch_choice = {'k': lambda: 'k-means++', 'r': lambda: 'random', 'm': lambda: mean_image(label_train, data_train)} kmeans = KMeans(n_clusters=10, random_state=0, init=switch_choice[args.init]()) \ .fit(extract_col(data_train, args.columns)) predicted = kmeans.predict(extract_col(data_test, args.columns)) print('kmeans done') compare_class(predicted, label_test) if args.create_mean: create_images_from_rows('km', mean_image(predicted, data_test))
for z in range(len(self.data[0])): """ class += (pval(y, z) - val(i, z) """ tmp[y] += (self.clusters[y].dimensions[z] - img[z])**2 max = tmp[0] id = 0 for y in range(self.nbclass): if tmp[y] < max: id = y return id if __name__ == '__main__': args = parse_args("kmeans").parse_args(["-r", "-s", "0.01"]) rand = np.random.randint(0, 10000000) print("Fetching data:") data, testdata = pre_processed_data(args, rand) print("Done") print("Fetching labels:") label, testlabel = pre_processed_label(args, rand) print("Done") kmeans = KMEANS(data, 10, label) print("Main Loop:") for i in range(10): print(i) print("E:") kmeans.E() print("Done") print("M:") kmeans.M() print("Done\n")
print('using all class') print('****%s****' % NB[name_nb]['name']) res = [] res.append( sk_bayes(NB[name_nb]['fn'], data_train, label_train, data_test, label_test, col)) for i in range(10): print('=======') label_train, label_test = fn_label(sep=SEP, i=i) print('class %d' % i) res.append( sk_bayes(NB[name_nb]['fn'], data_train, label_train, data_test, label_test, col)) if cm: for i in range(len(res)): create_images_from_rows('%d_b' % i, mean_image(res[i][1], res[i][0])) if __name__ == '__main__': args = bayes_args() rand = np.random.randint(10000000) data_train, data_test = pre_processed_data(args, rand, False) bayes(name_nb=args.bayes, fn_label=lambda sep='', i='': pre_processed_label( option=args, rand=rand, sep=sep, i=i), data_train=data_train, data_test=data_test, cm=args.create_mean, col=args.columns)