augustushsu / music-recommendation-system Goto Github PK

View Code? Open in Web Editor NEW

0.0 0.0 0.0 68.76 MB

Jupyter Notebook 64.40% Python 35.60%

music-recommendation-system's People

Contributors

Watchers

music-recommendation-system's Issues

Recommendation_Model.Accuracy is Repeated

accuracy(self) and self.accuracy is Repeated
So the accuracy is always 0 or 1
Figure out the difference between of tf.metrics.accuracy and the accuracy function

def accuracy(self):
    # Calculate accuracy
    with tf.name_scope('accuracy'):
        predictions = tf.argmax(self.Output, 1, name='predictions')
        correct_pred = tf.equal(predictions, tf.argmax(self.onehot_label, 1))
        self.accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

Loss Always Not Change

Use Softmax on the Output Layer, the loss is only decreased a little

with tf.name_scope('loss'):
    log_loss = tf.nn.softmax_cross_entropy_with_logits(logits=self.Output, labels=self.onehot_label)
    self.loss = tf.reduce_mean(log_loss)

with tf.name_scope('OUTPUT_LAYER'):
    initial_value = tf.truncated_normal([self.FC_output.shape[1].value, self.output_dim],
                                                               stddev=0.1)
    self.output_weight = tf.Variable(initial_value, name='output_weight')
    self.output_bais = tf.Variable(tf.zeros(shape=[self.output_dim]), name='output_bais')
    self.Output = tf.matmul(self.FC_output, self.output_weight) + self.output_bais
    self.Output = tf.nn.softmax(self.Output)
    print('Output')
    print(self.Output.shape)

Recommend Projects