python - Denoise Autoencoder test set got same results with different input -
i built denoise autoencoder example : https://github.com/aymericdamien/tensorflow-examples/blob/master/examples/3_neuralnetworks/autoencoder.py worked example
i changed nothing input data audio data wav song. input data coverted e.g. float32(batch_size, n_input), range (0,1). input data generated origin salt , pepper noise (that is: potion of data set 0, potion set 1; in test, 20% changed)
then use training, however, after lots of tests, noticed that: different test set data, trained model same result each sequence. while used np.random sequence test, result different test set result.
do has ideas what's possible wrong it? thanks
Comments
Post a Comment