AutoencoderNN-Classifier

Train your autoencoder:

$python test_autoencoder.py –d <#dataset>

Default Autoencoder layer-block:

-----functions.py--------

Encoder: Convolutional layer followed by Batch Normalizaiton. After the second normalization, there is a downsampling (Maxpooling). Dropout layers have been added after the second layer-block. Decoder: Decoder part is a mirror of the encoder with a sigmoid as an activation function for the last convolutional layer.

Train your classification model:

$python test_classification.py –d <#training set> –dl <#training labels> -t <#testset> -tl <#test labels> -model <#autoencoder model>

Taking the encoder part of the autoencoder and adds the following layers: ->Flatten layer ->Dense layer ->Batch Normalization ->Dropout layer ->Dense(units = 10)

Accuracy: 0,9952