A computer-implemented method of training an autoencoder to recover missing data is provided. The autoencoder includes an encoder for encoding its inputs into a latent space and a decoder for decoding the encodings from the latent space. The method comprises creating a first training set including a valid data set of multiple dimensions, and training the encoder and the decoder in a first training stage using the first training set to reduce a difference between the valid data set provided to the encoder and a data set decoded by the decoder. The method further comprises creating a second training set comprising an invalid data set, and training the encoder in a second training stage using the second training set to reduce a difference between encodings of valid data instances and encodings of their corresponding invalid data instances.