def main(argv): options, remainder = getopt.getopt(argv, "o:v", ["train_file="]) # Parse the arguments for opt, arg in options: if opt == "--train_file": train_file = arg # Featurize the train_file: global templated_data global y_list templated_data, y_list = featurize_file(train_file) # Initialize the sparse weights vector: global w_vec w_vec = np.zeros((1, n_features)) # w_vec[0,13] = 1 # w_vec[0,18] = 1 # w_vec[0,7] = 1 # w_vec[0,11] = 1 # w_vec[0,100]=45 # Train: # train(templated_data, hashed_feature_matrix) train(1, templated_data)
def main(argv): options, remainder = getopt.getopt(argv, 'o:v', ['train_file=',]) # Parse the arguments for opt, arg in options: if opt == '--train_file': train_file = arg # print train_file # Featurize the train_file: # templated_data, hashed_feature_matrix = featurize_file(train_file) global templated_data global y_list templated_data, y_list = featurize_file(train_file) # print templated_data # print y_list # exit() # Initialize the sparse weights vector: global w_vec # w_vec = lil_matrix((1,n_features)) # w_vec = np.ones((1,n_features)) # w_vec = np.full((1,n_features),0.1) w_vec = np.zeros((1,n_features)) # w_vec[0,13] = 1 # w_vec[0,18] = 1 # w_vec[0,7] = 1 # w_vec[0,11] = 1 # print w_vec # w_vec[0,100]=45 # w_vec = sparse.csr_matrix(w_vec) # exit() # Debug: # Sentence 0, token 1's features # print templated_data[0][1]['F'] # print templated_data[0][1] # Sentence 0, token 1's hashed features # print hashed_feature_matrix # exit() # Train: # train(templated_data, hashed_feature_matrix) train(templated_data) exit() # Train using the SGD: SGD(5, templated_data)