problem about model prediction

[rezazad68]

hello, at first, thank for your contribution and nice work

I got a problem about model prediction, just trained model with 10 epochs, after training model, when I used model.predict on whole validation set I got all same value for each sample of validation set (returned value for each class on each sample of validation is : 0.250, 0.251, 0.250, 0.250 ) does it right value? actually it assign any new sample to second class.

it would be great if u can help me with regards

[royaljain]

Too late I think but if someone else stumbles upon this, here's what solved this for me -

Initially the values are pretty low and THresholdRelu layer clips them back to 0 and hence we get same output for everything

[chaitjo]

@royaljain What did you do to solve the problem? A different activation function to the ThresholdReLU?