DeepMind Technologies Limited

Artificial intelligence-based image search refinement

Selecting reinforcement learning actions using a low-level controller

LEARNED COMPUTER CONTROL USING POINTING DEVICE AND KEYBOARD ACTIONS

Learning abstractions using patterns of activations of a neural network hidden layer

HIGH-PERFORMANCE AND LOW-COMPLEXITY NEURAL COMPRESSION FROM A SINGLE IMAGE, VIDEO OR AUDIO DATA

AUTO-REGRESSIVE NEURAL NETWORK SYSTEMS WITH A SOFT ATTENTION MECHANISM USING SUPPORT DATA PATCHES

Population based training of neural networks

GATED ATTENTION NEURAL NETWORKS

Action selection based on environment observations and textual instructions

Verifying the provenance of a digital object using watermarking and embeddings

AUGMENTING MACHINE LEARNING LANGUAGE MODELS USING SEARCH ENGINE RESULTS

Adaptive visual speech recognition

Beam search decoding with forward-looking scores

Training action-selection neural networks from demonstrations using multiple losses

Training action selection neural networks using look-ahead search

CONTINUOUS CONTROL WITH DEEP REINFORCEMENT LEARNING

Neural programming

Speech synthesis utilizing audio waveform difference signal(s)

Generating images using neural networks

Distributional reinforcement learning using quantile function neural networks

Distributed training using actor-critic reinforcement learning with off-policy correction factors

Fast exploration and learning of latent graph models

PROGRESSIVE NEURAL NETWORKS

Image processing of an environment to select an action to be performed by an agent interacting with the environment

DATA-EFFICIENT REINFORCEMENT LEARNING FOR CONTINUOUS CONTROL TASKS

Selecting actions from large discrete action sets using reinforcement learning

Graph neural network systems for generating structured representations of objects

DATA-DRIVEN ROBOT CONTROL

Reinforcement learning with scheduled auxiliary control

Action classification in video clips using attention-based neural networks

Selecting reinforcement learning actions using a low-level controller

Distributional reinforcement learning for continuous control tasks

TRAINING A SPEAKER NEURAL NETWORK USING ONE OR MORE LISTENER NEURAL NETWORKS

Learning abstractions using patterns of activations of a neural network hidden layer

Training machine learning models by determining update rules using neural networks

Rate control machine learning models with feedback control for video encoding

Classifying input examples using a comparison set

REINFORCEMENT AND IMITATION LEARNING FOR A TASK

Training a neural network using outputs of a corruption neural network

Augmented recurrent neural network with external memory

Population based training of neural networks

Learning observation representations by predicting the future in latent space

ALIGNING ENTITIES USING NEURAL NETWORKS

Reinforcement learning using distributed prioritized replay

Training more secure neural networks by using local linearity regularization

Multi-agent reinforcement learning with matchmaking policies

Augmenting machine learning language models using search engine results

Training neural networks using a prioritized experience memory

Learning from delayed outcomes using neural networks

PERFORMANCE OF A NEURAL NETWORK USING AUTOMATICALLY UNCOVERED FAILURE CASES

Dynamic placement of computation sub-graphs

Distributional reinforcement learning using quantile function neural networks

Parallel video processing systems

CONTROLLING INTERACTIVE AGENTS USING MULTI-MODAL INPUTS

Distributed training using actor-critic reinforcement learning with off-policy correction factors

Rating tasks and policies using conditional probability distributions derived from equilibrium-based solutions of games

Deep neural network system for similarity-based graph representations

Generating output examples using bit blocks

Learning visual concepts using neural networks

Training action selection neural networks using look-ahead search

Training action-selection neural networks from demonstrations using multiple losses

Autoregressive neural networks for the generation of polygonal meshes of 3D objects

RECURRENT UNIT FOR GENERATING OR PROCESSING A SEQUENCE OF IMAGES

Training action selection neural networks using apprenticeship

Generating images using neural networks

Recommending content using neural networks

Distributional reinforcement learning for continuous control tasks

Reinforcement learning for active sequence processing

Predicting spectral representations for training speech synthesis neural networks

PERSISTENT MESSAGE PASSING FOR GRAPH NEURAL NETWORKS

Resolving time-delays using generative models

Learning motor primitives and training a machine learning system using a linear-feedback-stabilized policy

Gated attention neural networks

Generating audio using neural networks

Speech coding using content latent embedding vectors and speaker latent embedding vectors

Action selection based on environment observations and textual instructions

Black-box optimization using neural networks

Iterative multiscale image generation using neural networks

Performing navigation tasks using grid codes

Reinforcement learning using baseline and policy neural networks

Speech synthesis utilizing audio waveform difference signal(s)

Generative adversarial networks with temporal and spatial discriminators for efficient video generation

Bandwidth extension of incoming data using neural networks

Unsupervised learning of object keypoint locations in images through temporal transport or spatio-temporal transport

Energy-based associative memory neural networks

Image processing with recurrent attention

Unsupervised control using learned rewards

Multi-task neural network systems with task-specific policies and a shared policy

Neural network systems for decomposing video data into layered representations

Neural network architecture for efficient resource allocation

MACHINE LEARNING FOR DETERMINING PROTEIN STRUCTURES

Pairwise ranking using neural networks

Selecting reinforcement learning actions using a low-level controller

PROTEIN STRUCTURE PREDICTION USING GEOMETRIC ATTENTION NEURAL NETWORKS

Generating audio data using unaligned text inputs with an adversarial network

NEURAL NETWORK OPTIMIZATION USING CURVATURE ESTIMATES BASED ON RECENT GRADIENTS

Recurrent neural networks for data item generation

Processing sequences using convolutional neural networks

Predicting properties of materials from physical material structures

Robustness to adversarial behavior for text classification models