When I train the transpose convolution, I always get zero accuracy.
this is my code
Please help me to solve this.
Hi @Owen_Grant, Here the use case you are trying to solve is to reconstruct the given input, where the output should closely approximate the input and the loss function you are using for this task is mean_squared_error which is suitable for regression task, in this case accuracy will not be a suitable metric because it is generally designed for classification tasks, where exact match between predicted and true labels are counted as correct. Thank You.
Thanks.
Then which loss function should I use?
Hi @Owen_Grant, In this case you can use the mse
for the metrics. Thank You.