Autoencoders now available

The self-supervised tasks of autoencoders and variational autoencoders join the supervised tasks of regression, binary/multiclass/multilabel classification, and survival available at Alea Axis.

Autoencoders are a powerful method of dimensionality reduction and are a common tool in many researchers’ toolbox. Now they are available to anyone with an internet connection.

Simply head over to aleaaxis.net/train/autoencoder and build your first encoder and decoder with your own data or the provided MNIST data. If you set K-folds to 1 you will be able to view the decoder’s reconstruction of each latent vector at the completion of training.