diff --git a/Brain_Integ.mat b/Brain_Integ.mat new file mode 100755 index 0000000..5a5aee6 Binary files /dev/null and b/Brain_Integ.mat differ diff --git a/reg.py b/reg.py new file mode 100644 index 0000000..4ae8f27 --- /dev/null +++ b/reg.py @@ -0,0 +1,161 @@ +from __future__ import absolute_import, division, print_function + +import tensorflow as tf + +# import matplotlib.pyplot as plt +import numpy as np +import scipy.io as sio +from sklearn.utils import shuffle +from sklearn.model_selection import train_test_split + +def read_dataset(): + + D = sio.loadmat('Brain_Integ.mat') + # Extract the fearues and labels + X = D['Integ_X'].astype('float32') # 560*399 + T = np.asarray([t[0] for t in D['Survival']]).astype('float32') # 560*1 + C = np.asarray([c[0] for c in D['Censored']]).astype('int32') # O # 560*1 + # TC = np.asarray([T for c in C if c==1]) + # Consider only the dead patient's data + TC = T[C==1] + XC = X[C==1] + Y = np.asarray(TC) + X = np.asarray(XC) + + print ('Shape of X : ', X.shape) + print ('Shape of Y : ', Y.shape) + return (X, Y) + +X, Y = read_dataset() +X, Y = shuffle(X, Y, random_state=1) + +X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.20, random_state=420) + + +total_len = X_train.shape[0] # Total Samples + +# ***** Parameters ***** +learning_rate = 0.01 # Learning rate for parameter updates +training_epochs = 1000 # EPOCHS +batch_size = 100 # Set to 1 for whole samples at once ## THIS CAUSES ERROR IN label = batch_y +display_step = 100 # Set to 1 for displaying all epochs +# dropout_rate = 0.9 + +# ***** Network Parameters ***** +n_hidden_1 = 500 # 1st layer number of neurons +n_hidden_2 = 500 # 2nd layer number of neurons +n_hidden_3 = 500 # 3rd layer number of neurons +n_hidden_4 = 500 # 4th layer number of neurons +n_input = X_train.shape[1] # number of columns(features) +n_classes = 1 # Only one output column, represents the predicted value + +# tf Graph input +x = tf.placeholder("float", [None, None]) # Set to [, None] to definite value +y = tf.placeholder("float", [None]) # Set to a definite value + +# layers' weights & biases initialization +variance = 0.1 # VARIANCE selection highly affects ## If set to 10, results in NaN values for predication and cost +weights = { + 'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1], 0, variance)), + 'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2], 0, variance)), + 'h3': tf.Variable(tf.random_normal([n_hidden_2, n_hidden_3], 0, variance)), + 'h4': tf.Variable(tf.random_normal([n_hidden_3, n_hidden_4], 0, variance)), + 'out': tf.Variable(tf.random_normal([n_hidden_4, n_classes], 0, variance)) +} +biases = { + 'b1': tf.Variable(tf.random_normal([n_hidden_1], 0, variance)), + 'b2': tf.Variable(tf.random_normal([n_hidden_2], 0, variance)), + 'b3': tf.Variable(tf.random_normal([n_hidden_3], 0, variance)), + 'b4': tf.Variable(tf.random_normal([n_hidden_4], 0, variance)), + 'out': tf.Variable(tf.random_normal([n_classes], 0, variance)) +} + +# print (weights['h1']) +# print (biases['b1']) +# Model Creation +def multilayer_perceptron(x, weights, biases): + # Hidden layers with RELU activation + layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1']) + layer_1 = tf.nn.relu(layer_1) + + layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2']) + layer_2 = tf.nn.relu(layer_2) + + layer_3 = tf.add(tf.matmul(layer_2, weights['h3']), biases['b3']) + layer_3 = tf.nn.relu(layer_3) + + layer_4 = tf.add(tf.matmul(layer_3, weights['h4']), biases['b4']) + layer_4 = tf.nn.relu(layer_4) + + out_layer = tf.matmul(layer_4, weights['out']) + biases['out'] + return out_layer + + +# Construct the grpah for forward pass +pred = multilayer_perceptron(x, weights, biases) + +# Defining loss and optimizer +### reduce_sum ### +# cost = (tf.reduce_mean(tf.square(tf.transpose(pred)-y)) + 0.01*tf.nn.l2_loss(weights)) # USE tf.transpose ====> those 2 arrays are row and column, results in unexpected format, if let as early +cost = tf.reduce_mean(tf.square(tf.transpose(pred)-y)) +optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost) # Change to tf.GradientDescentOptimizer() and see + + + + +# Launch the graph +with tf.Session() as sess: + + # Select either Initializers + sess.run(tf.initialize_all_variables()) + # tf.global_variables_initializer() + # ============================= Each EPOCH graph Session============================================================== + + for epoch in range(training_epochs): + avg_cost = 0.0 + total_batch = int(total_len/batch_size) + # Loop over all batches + for i in range(total_batch-1): + batch_x = X_train[i*batch_size:(i+1)*batch_size] + batch_y = Y_train[i*batch_size:(i+1)*batch_size] + print ("batch number :", i) + # Run optimization op (backprop) and cost op (to get loss value) + _, c, p = sess.run([optimizer, cost, pred], feed_dict={x: batch_x, y: batch_y}) + # Compute average loss + avg_cost += c / total_batch + + + # sample prediction + label_value = batch_y + estimate = p # Predicted value + err = label_value-estimate + # print ("num batch:", total_batch) + + # Display logs per epoch step + if epoch % display_step == 0: + print ("Epoch:", '%4d' % (epoch+1), "cost=", "{:.9f}".format(avg_cost)) + print ("[*]----------------------------") + for i in xrange(3): # Comparing the predicted values with Labeled values in display + print ("label value:", label_value[i], "estimated value:", estimate[i]) + print ("[*]============================") + + # ==================================================================================================================== + print ("Optimization Finished!") + + # # Test model + # correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1)) + # # Calculate accuracy + # accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float")) + # print ("Accuracy:", accuracy.eval({x: X_test, y: Y_test})) + + print ("TESTING STARTS") + + accuracy = sess.run(cost, feed_dict={x:X_test, y: Y_test}) + predicted_vals = sess.run(pred, feed_dict={x: X_test}) + # print ("Accuracy:", accuracy, "Predicted VALUE : ", predicted_vals) + print ("Labeled value : ", Y_test, " <<<>>> Predicted VALUE : ", predicted_vals) + + print ("TESTING ENDS") + + +