Learning observation representations by predicting the future in latent space

A system comprising:

• one or more computers;
• and
• one or more storage devices communicatively coupled to the one or more computers, wherein the one or more storage devices store instructions that, when executed by the one or more computers, cause the one or more computers to perform operations comprising:
• receiving an input observation in a data space;
• and
• processing the input observation using an encoder neural network, in accordance with a plurality of encoder neural network parameters, to generate a latent representation of the input observation in a latent space;
• wherein the encoder neural network has been trained to optimize a loss function, wherein, for a first observation and a separate second observation that is different from the first observation, the loss function measures an error between:
• (i) a latent representation of the second observation in the latent space, wherein the latent representation of the second observation is generated by processing the second observation using the encoder neural network;
• and
• (ii) a prediction for the latent representation of the second observation in the latent space;
• wherein the prediction for the latent representation of the second observation is generated based on a latent representation of the first observation,
• wherein the latent representation of the first observation is generated by processing the first observation using the encoder neural network.

One or more non-transitory computer storage media storing instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:

• receiving an input observation in a data space;
• and
• processing the input observation using an encoder neural network, in accordance with a plurality of encoder neural network parameters, to generate a latent representation of the input observation in a latent space;
• wherein the encoder neural network has been trained to optimize a loss function, wherein, for a first observation and a separate second observation that is different from the first observation, the loss function measures an error between:
• (i) a latent representation of the second observation in the latent space, wherein the latent representation of the second observation is generated by processing the second observation using the encoder neural network;
• and
• (ii) a prediction for the latent representation of the second observation in the latent space;
• wherein the prediction for the latent representation of the second observation is generated based on a latent representation of the first observation,
• wherein the latent representation of the first observation is generated by processing the first observation using the encoder neural network.