from sklearn.metrics import confusion_matrix #performance diagnostic tool import pdb #debugging package - use by including 'pdb.set_trace()' in code #### LOAD DATA #### setnum = 1 #1 is opp, 2 is pam, 3 is skoda if setnum == 1: DB = 79 #reference to the data set to load below if setnum == 2: DB = 52 if setnum == 3: DB = 60 #load the database through the function 'loadingDB()' from dataset.py train_x, valid_x, test_x, train_y, valid_y, test_y = loadingDB('../', DB) #csv = open(str(DB)+'.csv','a') #to store performance results later on #### FUNCTION THAT DETERMINES THAT DATA SET USED IN EACH ITERATION #### def create_batches(data_x,data_y,range_split, random_start = False): dim_data = train_x.shape[1] n_classes = train_y.shape[1] #determine number of batches in range (min_n_batches:max_n_batches) n_batches = np.random.randint(range_split[0],range_split[1],1)[0] #use [0] because the function returns an array and we want a number only #the length of each batch l_batches = data_x.shape[0]//n_batches
import tensorflow as tf from tensorflow.contrib import rnn from dataset import loadingDB import numpy as np from sklearn.metrics import f1_score from sklearn.metrics import confusion_matrix import pickle import pandas as pd import pdb #debugging package - use by including 'pdb.set_trace()' in code ## this is a test function #R select the data set to use an d load data setnum = 1#1 is opp79, 2 is pamap2, 3 is skoda if setnum == 1: train_x, valid_x, test_x, train_y, valid_y, test_y = loadingDB('../', 79) n_classes = 18 DB = 79 #number of features if setnum == 2: train_x, valid_x, test_x, train_y, valid_y, test_y = loadingDB('../', 52) n_classes = 12 DB = 52 if setnum == 3: train_x, valid_x, test_x, train_y, valid_y, test_y = loadingDB('../', 60) n_classes = 11 DB = 60 # set hyperparameters of structure, dim and DB are same nm_epochs = 2 rnn_size = 256 #number of nodes in the hidden layer number_of_layers = 2 keep_rate = 0.5 #1/dropout rate