0

I'm building convolutional neural network for classification of the data into different categories The input data is of shape : 30000, 6, 15, 1 the data has 30000 samples, 15 predictors and 6 possible categories.

The weight and bias dictionary I'm using is as follows:

weights = {
    'wc1': tf.get_variable('W0', shape=(3,3,1,8), initializer=tf.contrib.layers.xavier_initializer()), 
    'wc2': tf.get_variable('W1', shape=(3,3,8,12), initializer=tf.contrib.layers.xavier_initializer()), 
    'wc3': tf.get_variable('W2', shape=(3,3,12,16), initializer=tf.contrib.layers.xavier_initializer()), 
    'wc4': tf.get_variable('W3', shape=(3,3,16,20), initializer=tf.contrib.layers.xavier_initializer()),
    'wd1': tf.get_variable('W4', shape=(4*4*20,20), initializer=tf.contrib.layers.xavier_initializer()), 
    'out': tf.get_variable('W6', shape=(20,n_classes), initializer=tf.contrib.layers.xavier_initializer()), 
}


biases = {
    'bc1': tf.get_variable('B0', shape=(8), initializer=tf.contrib.layers.xavier_initializer()),
    'bc2': tf.get_variable('B1', shape=(12), initializer=tf.contrib.layers.xavier_initializer()),
    'bc3': tf.get_variable('B2', shape=(16), initializer=tf.contrib.layers.xavier_initializer()),
    'bc4': tf.get_variable('B3', shape=(20), initializer=tf.contrib.layers.xavier_initializer()),
    'bd1': tf.get_variable('B4', shape=(20), initializer=tf.contrib.layers.xavier_initializer()),
    'out': tf.get_variable('B5', shape=(6), initializer=tf.contrib.layers.xavier_initializer()),
}

As expected the output tensor of :

def conv_net(x, weights, biases):  
    conv1 = conv2d(x, weights['wc1'], biases['bc1'])
    conv1 = maxpool2d(conv1, k=2)

    conv2 = conv2d(conv1, weights['wc2'], biases['bc2'])
    conv2 = maxpool2d(conv2, k=2)

    conv3 = conv2d(conv2, weights['wc3'], biases['bc3'])
    conv3 = maxpool2d(conv3, k=2)

    conv4 = conv2d(conv3, weights['wc4'], biases['bc4'])
    conv4 = maxpool2d(conv4, k=2)


    # Fully connected layer
    # Reshape conv2 output to fit fully connected layer input
    fc1 = tf.reshape(conv4, [-1, weights['wd1'].get_shape().as_list()[0]])
    fc1 = tf.add(tf.matmul(fc1, weights['wd1']), biases['bd1'])
    fc1 = tf.nn.relu(fc1)
    # Output, class prediction 
    out = tf.add(tf.matmul(fc1, weights['out']), biases['out'])
    return out

When x = x(batch size = 64) is of shape (4,6).

But since the labels for the batch os 64 is of shape [64,6] where 6 is the number of categories, the cost function defined as 

cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y))

where pred = conv_net(x, weights, biases)

gives the error : 

InvalidArgumentError (see above for traceback): logits and labels must be broadcastable: logits_size=[4,6] labels_size=[64,6]
     [[Node: softmax_cross_entropy_with_logits_sg = SoftmaxCrossEntropyWithLogits[T=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:CPU:0"](Add_1, softmax_cross_entropy_with_logits_sg/Reshape_1)]]

If my understanding is correct it has to do with the definition of fully connected layer and output layer filter size in weights library. Am I understanding right and if yes, what should be the filter shape in FC and output layers? and what's the underlying logic?

iprof0214
  • 701
  • 2
  • 6
  • 19
  • Why do you expect the output tensor of `conv_net` to be of shape (4, 6)? I expect the output tensor, or `pred`, to be (64, 6) because the first dimension means batch size and the second does the number of categories. Furthermore, what do you mean with x = x(batch size = 64)?` – T. Ogawa Nov 07 '18 at 20:39
  • x = x(batch size = 64) : I mean x has 64 samples. I expect the output od conv_net to be 4,6 instead of 64,6 as the weights of the dense layer/fully connected layer is defined as 4*4*20. I', having a confusion understanding the output shape of the conv layers here. I'd appreciate an input – iprof0214 Nov 07 '18 at 21:07
  • Why when the weights of the dense layer/fully connected layer is defined as 4*4*20, do you expect the output of conversion_net to be 4, 6? What is 4? – T. Ogawa Nov 07 '18 at 21:13
  • Thanks, I calculated the output shapes from each conv layer and the output was 1*1. I changed the dense layer flatten parameters to 1*1*20 and it worked. – iprof0214 Nov 08 '18 at 14:37

0 Answers0