Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network that is used to select actions to be performed by a reinforcement learning agent interacting with an environment. In particular, the actions are selected from a continuous action space and the system trains the action selection neural network jointly with a distribution Q network that is used to update the parameters of the action selection neural network.

Abstract: An off-policy reinforcement learning actor-critic neural network system configured to select actions from a continuous action space to be performed by an agent interacting with an environment to perform a task. An observation defines environment state data and reward data. The system has an actor neural network which learns a policy function mapping the state data to action data. A critic neural network learns an action-value (Q) function. A replay buffer stores tuples of the state data, the action data, the reward data and new state data. The replay buffer also includes demonstration transition data comprising a set of the tuples from a demonstration of the task within the environment. The neural network system is configured to train the actor neural network and the critic neural network off-policy using stored tuples from the replay buffer comprising tuples both from operation of the system and from the demonstration transition data.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating output examples using neural networks. One of the methods includes receiving a request to generate an output example of a particular type, accessing dependency data, and generating the output example by, at each of a plurality of generation time steps: identifying one or more current blocks for the generation time step, wherein each current block is a block for which the values of the bits in all of the other blocks identified in the dependency for the block have already been generated; and generating the values of the bits in the current blocks for the generation time step conditioned on, for each current block, the already generated values of the bits in the other blocks identified in the dependency for the current block.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating images using neural networks. One of the methods includes generating the output image pixel by pixel from a sequence of pixels taken from the output image, comprising, for each pixel in the output image, generating a respective score distribution over a discrete set of possible color values for each of the plurality of color channels.

Abstract: Methods, systems and apparatus, including computer programs encoded on computer storage media, for training an action selection neural network. One of the methods includes receiving an observation characterizing a current state of the environment; determining a target network output for the observation by performing a look ahead search of possible future states of the environment starting from the current state until the environment reaches a possible future state that satisfies one or more termination criteria, wherein the look ahead search is guided by the neural network in accordance with current values of the network parameters; selecting an action to be performed by the agent in response to the observation using the target network output generated by performing the look ahead search; and storing, in an exploration history data store, the target network output in association with the observation for use in updating the current values of the network parameters.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a plurality of possible segmentations of an image. In one aspect, a method comprises: receiving a request to generate a plurality of possible segmentations of an image; sampling a plurality of latent variables from a latent space, wherein each latent variable is sampled from the latent space in accordance with a respective probability distribution over the latent space that is determined based on the image; generating a plurality of possible segmentations of the image, comprising, for each latent variable, processing the image and the latent variable using a segmentation neural network having a plurality of segmentation neural network parameters to generate the possible segmentation of the image; and providing the plurality of possible segmentations of the image in response to the request.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for evaluating reinforcement learning policies. One of the methods includes receiving a plurality of training histories for a reinforcement learning agent; determining a total reward for each training observation in the training histories; partitioning the training observations into a plurality of partitions; determining, for each partition and from the partitioned training observations, a probability that the reinforcement learning agent will receive the total reward for the partition if the reinforcement learning agent performs the action for the partition in response to receiving the current observation; determining, from the probabilities and for each total reward, a respective estimated value of performing each action in response to receiving the current observation; and selecting an action from the pre-determined set of actions from the estimated values in accordance with an action selection policy.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a target sequence from an input sequence. In one aspect, a method comprises maintaining a set of current hypotheses, wherein each current hypothesis comprises an input prefix and an output prefix. For each possible combination of input and output prefix length, the method extends any current hypothesis that could reach the possible combination to generate respective extended hypotheses for each such current hypothesis; determines a respective direct score for each extended hypothesis using a direct model; determines a first number of highest-scoring hypotheses according to the direct scores; rescores the first number of highest-scoring hypotheses using a noisy channel model to generate a reduced number of hypotheses; and adds the reduced number of hypotheses to the set of current hypotheses.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a system output using a remembered value of a neural network hidden state. In one aspect, a system comprises an external memory that maintains context experience tuples respectively comprising: (i) a key embedding of context data, and (ii) a value of a hidden state of a neural network at the respective previous time step. The neural network is configured to receive a system input and a remembered value of the hidden state of the neural network and to generate a system output. The system comprises a memory interface subsystem that is configured to determine a key embedding for current context data, determine a remembered value of the hidden state of the neural network based on the key embedding, and provide the remembered value of the hidden state as an input to the neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating audio output samples predicted to be communicated by a user. One example system includes a first user device having a first user. The first user device initiates a communication session between the first user and a second user of a second user device. The first user device obtains a neural network model of the second user. The neural network model is trained to generate, conditioned on audio input samples received up to a current time step, an audio output sample predicted to be communicated by the second user at a next time step. The user device repeatedly provides received audio input samples as input to the neural network model and plays audio output samples generated by the neural network model in place of received audio input samples communicated by the second user.

Abstract: A computer-implemented method of training a student machine learning system comprises receiving data indicating execution of an expert, determining one or more actions performed by the expert during the execution and a corresponding state-action Jacobian, and training the student machine learning system using a linear-feedback-stabilized policy. The linear-feedback-stabilized policy may be based on the state-action Jacobian. Also a neural network system for representing a space of probabilistic motor primitives, implemented by one or more computers. The neural network system comprises an encoder configured to generate latent variables based on a plurality of inputs, each input comprising a plurality of frames, and a decoder configured to generate an action based on one or more of the latent variables and a state.

Abstract: A system implemented by one or more computers comprises a visual encoder component configured to receive as input data representing a sequence of image frames, in particular representing objects in a scene of the sequence, and to output a sequence of corresponding state codes, each state code comprising vectors, one for each of the objects. Each vector represents a respective position and velocity of its corresponding object. The system also comprises a dynamic predictor component configured to take as input a sequence of state codes, for example from the visual encoder, and predict a state code for a next unobserved frame. The system further comprises a state decoder component configured to convert the predicted state code, to a state, the state comprising a respective position and velocity vector for each object in the scene. This state may represent a predicted position and velocity vector for each of the objects.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing visual speech recognition. In one aspect, a method comprises receiving a video comprising a plurality of video frames, wherein each video frame depicts a pair of lips; processing the video using a visual speech recognition neural network to generate, for each output position in an output sequence, a respective output score for each token in a vocabulary of possible tokens, wherein the visual speech recognition neural network comprises one or more volumetric convolutional neural network layers and one or more time-aggregation neural network layers; wherein the vocabulary of possible tokens comprises a plurality of phonemes; and determining a sequence of words expressed by the pair of lips depicted in the video using the output scores.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating an output sequence of audio data that comprises a respective audio sample at each of a plurality of time steps. One of the methods includes, for each of the time steps: providing a current sequence of audio data as input to a convolutional subnetwork, wherein the current sequence comprises the respective audio sample at each time step that precedes the time step in the output sequence, and wherein the convolutional subnetwork is configured to process the current sequence of audio data to generate an alternative representation for the time step; and providing the alternative representation for the time step as input to an output layer, wherein the output layer is configured to: process the alternative representation to generate an output that defines a score distribution over a plurality of possible audio samples for the time step.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for selecting actions to be performed by an agent interacting with an environment. In one aspect, a system comprises a grid cell neural network and an action selection neural network. The grid cell network is configured to: receive an input comprising data characterizing a velocity of the agent; process the input to generate a grid cell representation; and process the grid cell representation to generate an estimate of a position of the agent in the environment; the action selection neural network is configured to: receive an input comprising a grid cell representation and an observation characterizing a state of the environment; and process the input to generate an action selection network output.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing video data. An example system receives video data and generates optical flow data. An image sequence from the video data is provided to a first 3D spatio-temporal convolutional neural network to process the image data in at least three space-time dimensions and to provide a first convolutional neural network output. A corresponding sequence of optical flow image frames is provided to a second 3D spatio-temporal convolutional neural network to process the optical flow data in at least three space-time dimensions and to provide a second convolutional neural network output. The first and second convolutional neural network outputs are combined to provide a system output.

Abstract: A method of generating an output image having an output resolution of N pixels×N pixels, each pixel in the output image having a respective color value for each of a plurality of color channels, the method comprising: obtaining a low-resolution version of the output image; and upscaling the low-resolution version of the output image to generate the output image having the output resolution by repeatedly performing the following operations: obtaining a current version of the output image having a current K×K resolution; and processing the current version of the output image using a set of convolutional neural networks that are specific to the current resolution to generate an updated version of the output image having a 2K×2K resolution.

Abstract: Methods and systems for determining an optimized setting for one or more process parameters of a machine learning training process are described. One of the methods includes processing a current network input using a recurrent neural network in accordance with first values of the network parameters to obtain a current network output, obtaining a measure of the performance of the machine learning training process with an updated setting defined by the current network output, and generating a new network input that includes (i) the updated setting defined by the current network output and (ii) the measure of the performance of the training process with the updated setting defined by the current network output.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for learning visual concepts using neural networks. One of the methods includes receiving a new symbol input comprising one or more symbols from a vocabulary; and generating a new output image that depicts concepts referred to by the new symbol input, comprising: processing the new symbol input using a symbol encoder neural network to generate a new symbol encoder output for the new symbol input; sampling, from the distribution parameterized by the new symbol encoder output, a respective value for each of a plurality of visual factors; and processing a new image decoder input comprising the respective values for the visual factors using an image decoder neural network to generate the new output image.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing images using recurrent attention. One of the methods includes determining a location in the first image; extracting a glimpse from the first image using the location; generating a glimpse representation of the extracted glimpse; processing the glimpse representation using a recurrent neural network to update a current internal state of the recurrent neural network to generate a new internal state; processing the new internal state to select a location in a next image in the image sequence after the first image; and processing the new internal state to select an action from a predetermined set of possible actions.