Abstract: Systems, methods, devices, and other techniques are disclosed for using an augmented neural network system to generate a sequence of outputs from a sequence of inputs. An augmented neural network system can include a controller neural network, a hierarchical external memory, and a memory access subsystem. The controller neural network receives a neural network input at each of a series of time steps processes the neural network input to generate a memory key for the time step. The external memory includes a set of memory nodes arranged as a binary tree. To provide an interface between the controller neural network and the external memory, the system includes a memory access subsystem that is configured to, for each of the series of time steps, perform one or more operations to generate a respective output for the time step. The capacity of the neural network system to account for long-range dependencies in input sequences may be extended.

Abstract: Systems, methods, and apparatus, including computer programs encoded on a computer storage medium, related to associative long short-term memory (LSTM) neural network layers configured to maintain N copies of an internal state for the associative LSTM layer, N being an integer greater than one. In one aspect, a system includes a recurrent neural network including an associative LSTM layer, wherein the associative LSTM layer is configured to, for each time step, receive a layer input, update each of the N copies of the internal state using the layer input for the time step and a layer output generated by the associative LSTM layer for a preceding time step, and generate a layer output for the time step using the N updated copies of the internal state.

Abstract: Methods, systems, and apparatus for classifying a new example using a comparison set of comparison examples. One method includes maintaining a comparison set, the comparison set including comparison examples and a respective label vector for each of the comparison examples, each label vector including a respective score for each label in a predetermined set of labels; receiving a new example; determining a respective attention weight for each comparison example by applying a neural network attention mechanism to the new example and to the comparison examples; and generating a respective label score for each label in the predetermined set of labels from, for each of the comparison examples, the respective attention weight for the comparison example and the respective label vector for the comparison example, in which the respective label score for each of the labels represents a likelihood that the label is a correct label for the new example.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a reinforcement learning system. The method includes: training an action selection policy neural network, and during the training of the action selection neural network, training one or more auxiliary control neural networks and a reward prediction neural network. Each of the auxiliary control neural networks is configured to receive a respective intermediate output generated by the action selection policy neural network and generate a policy output for a corresponding auxiliary control task. The reward prediction neural network is configured to receive one or more intermediate outputs generated by the action selection policy neural network and generate a corresponding predicted reward.

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 and systems for performing a sequence of machine learning tasks. One system includes a sequence of deep neural networks (DNNs), including: a first DNN corresponding to a first machine learning task, wherein the first DNN comprises a first plurality of indexed layers, and each layer in the first plurality of indexed layers is configured to receive a respective layer input and process the layer input to generate a respective layer output; and one or more subsequent DNNs corresponding to one or more respective machine learning tasks, wherein each subsequent DNN comprises a respective plurality of indexed layers, and each layer in a respective plurality of indexed layers with index greater than one receives input from a preceding layer of the respective subsequent DNN, and one or more preceding layers of respective preceding DNNs, wherein a preceding layer is a layer whose index is one less than the current index.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a machine learning model. In one aspect, a method includes receiving training data for training the machine learning model on a plurality of tasks, where each task includes multiple batches of training data. A task is selected in accordance with a current task selection policy. A batch of training data is selected from the selected task. The machine learning model is trained on the selected batch of training data to determine updated values of the model parameters. A learning progress measure that represents a progress of the training of the machine learning model as a result of training the machine learning model on the selected batch of training data is determined. The current task selection policy is updated using the learning progress measure.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for asynchronous deep reinforcement learning. One of the systems includes a plurality of workers, wherein each worker is configured to operate independently of each other worker, and wherein each worker is associated with a respective actor that interacts with a respective replica of the environment during the training of the deep neural network.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for augmenting neural networks with an external memory. One of the systems includes a controller neural network that includes a Least Recently Used Access (LRUA) subsystem configured to: maintain a respective usage weight for each of a plurality of locations in the external memory, and for each of the plurality of time steps: generate a respective reading weight for each location using a read key, read data from the locations in accordance with the reading weights, generate a respective writing weight for each of the locations from a respective reading weight from a preceding time step and the respective usage weight for the location, write a write vector to the locations in accordance with the writing weights, and update the respective usage weight from the respective reading weight and the respective writing weight.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for selecting actions from large discrete action sets. One of the methods includes receiving a particular observation representing a particular state of an environment; and selecting an action from a discrete set of actions to be performed by an agent interacting with the environment, comprising: processing the particular observation using an actor policy network to generate an ideal point; determining, from the points that represent actions in the set, the k nearest points to the ideal point; for each nearest point of the k nearest points: processing the nearest point and the particular observation using a Q network to generate a respective Q value for the action represented by the nearest point; and selecting the action to be performed by the agent from the k actions represented by the k nearest points based on the Q values.

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 training an action selection policy neural network. In one aspect, a method comprises: obtaining an expert observation; processing the expert observation using a generative neural network system to generate a given observation-given action pair, wherein the generative neural network system has been trained to be more likely to generate a particular observation-particular action pair if performing the particular action in response to the particular observation is more likely to result in the environment later reaching the state characterized by a target observation; processing the given observation using the action selection policy neural network to generate a given action score for the given action; and adjusting the current values of the action selection policy neural network parameters to increase the given action score for the given action.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating sequences of predicted observations, for example images. In one aspect, a system comprises a controller recurrent neural network, and a decoder neural network to process a set of latent variables to generate an observation. An external memory and a memory interface subsystem is configured to, for each of a plurality of time steps, receive an updated hidden state from the controller, generate a memory context vector by reading data from the external memory using the updated hidden state, determine a set of latent variables from the memory context vector, generate a predicted observation by providing the set of latent variables to the decoder neural network, write data to the external memory using the latent variables, the updated hidden state, or both, and generate a controller input for a subsequent time step from the latent variables.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for reinforcement learning. One of the methods includes selecting an action to be performed by the agent using both a slow updating recurrent neural network and a fast updating recurrent neural network that receives a fast updating input that includes the hidden state of the slow updating recurrent neural network.

Abstract: Methods, systems and apparatus, including computer programs encoded on computer storage media, for training a value neural network that is configured to receive an observation characterizing a state of an environment being interacted with by an agent and to process the observation in accordance with parameters of the value neural network to generate a value score. One of the systems performs operations that include training a supervised learning policy neural network; initializing initial values of parameters of a reinforcement learning policy neural network having a same architecture as the supervised learning policy network to the trained values of the parameters of the supervised learning policy neural network; training the reinforcement learning policy neural network on second training data; and training the value neural network to generate a value score for the state of the environment that represents a predicted long-term reward resulting from the environment being in the state.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for reinforcement learning. The embodiments described herein apply meta-learning (and in particular, meta-gradient reinforcement learning) to learn an optimum return function G so that the training of the system is improved. This provides a more effective and efficient means of training a reinforcement learning system as the system is able to converge on an optimum set of one or more policy parameters ? more quickly by training the return function G as it goes. In particular, the return function G is made dependent on the one or more policy parameters ? and a meta-objective function J? is used that is differentiated with respect to the one or more return parameters ? to improve the training of the return function G.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating data items. One of the systems is a neural network system comprising a memory storing a plurality of template data items; one or more processors configured to select a memory address based upon a received input data item, and retrieve a template data item from the memory based upon the selected memory address; an encoder neural network configured to process the received input data item and the retrieved template data item to generate a latent variable representation; and a decoder neural network configured to process the retrieved template data item and the latent variable representation to generate an output data item.

Abstract: A neural network system is proposed to select actions to be performed by an agent interacting with an environment to perform a task in an attempt to achieve a specified result. The system may include a controller to receive state data and context data, and to output action data. The system may also include an imagination module to receive the state and action data, and to output consequent state data. The system may also include a manager to receive the state data and the context data, and to output route data which defines whether the system is to execute an action or to imagine. The system may also include a memory to store the context data.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for controlling an agent interacting with an environment. One of the methods includes obtaining a representation of an observation; processing the representation using a convolutional long short-term memory (LSTM) neural network comprising a plurality of convolutional LSTM neural network layers; processing an action selection input comprising the final LSTM hidden state output for the time step using an action selection neural network that is configured to receive the action selection input and to process the action selection input to generate an action selection output that defines an action to be performed by the agent at the time step; selecting, from the action selection output, the action to be performed by the agent at the time step in accordance with an action selection policy; and causing the agent to perform the selected action.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for executing a signal generation neural network on parallel processing hardware. One of the methods includes receiving weight matrices of a layer of a signal generation neural network. Rows of a first matrix for the layer are interleaved by assigning groups of rows of the first matrix to respective thread blocks of a plurality of thread blocks. A first subset of rows of the one or more other weight matrices are assigned to a first subset of the plurality of thread blocks and a second subset of rows of the one or more other weight matrices are assigned to a second subset of the plurality of thread blocks. The first matrix operation is performed substantially in parallel by the plurality of thread blocks. The other matrix operations are performed substantially in parallel by the plurality of thread blocks.