公开(公告)号：US10789511B2

公开(公告)日：2020-09-29

申请号：US16601324

申请日：2019-10-14

Applicant: DeepMind Technologies Limited

Inventor： Gregory Duncan Wayne ,  Timothy Paul Lillicrap ,  Chia-Chun Hung ,  Joshua Simon Abramson

IPC: G06K9/62 ,  G06F11/30 ,  G06N3/08

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment to perform a specified task. One of the methods includes causing the agent to perform a task episode in which the agent attempts to perform the specified task; for each of one or more particular time steps in the sequence: generating a modified reward for the particular time step from (i) the actual reward at the time step and (ii) value predictions at one or more time steps that are more than a threshold number of time steps after the particular time step in the sequence; and training, through reinforcement learning, the neural network system using at least the modified rewards for the particular time steps.

公开(公告)号：US20210081723A1

公开(公告)日：2021-03-18

申请号：US17035546

申请日：2020-09-28

Applicant: DeepMind Technologies Limited

Inventor： Gregory Duncan Wayne ,  Timothy Paul Lillicrap ,  Chia-Chun Hung ,  Joshua Simon Abramson

IPC: G06K9/62 ,  G06F11/30 ,  G06N3/08

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment to perform a specified task. One of the methods includes causing the agent to perform a task episode in which the agent attempts to perform the specified task; for each of one or more particular time steps in the sequence: generating a modified reward for the particular time step from (i) the actual reward at the time step and (ii) value predictions at one or more time steps that are more than a threshold number of time steps after the particular time step in the sequence; and training, through reinforcement learning, the neural network system using at least the modified rewards for the particular time steps.

公开(公告)号：US11769049B2

公开(公告)日：2023-09-26

申请号：US17035546

申请日：2020-09-28

Applicant: DeepMind Technologies Limited

Inventor： Gregory Duncan Wayne ,  Timothy Paul Lillicrap ,  Chia-Chun Hung ,  Joshua Simon Abramson

IPC: G06K9/62 ,  G06F11/30 ,  G06N3/08 ,  G06F18/21 ,  G06V10/764 ,  G06V10/774 ,  G06V10/778 ,  G06V10/82

CPC classification number: G06N3/08 ,  G06F11/3037 ,  G06F11/3072 ,  G06F18/2193 ,  G06V10/764 ,  G06V10/774 ,  G06V10/7796 ,  G06V10/82

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment to perform a specified task. One of the methods includes causing the agent to perform a task episode in which the agent attempts to perform the specified task; for each of one or more particular time steps in the sequence: generating a modified reward for the particular time step from (i) the actual reward at the time step and (ii) value predictions at one or more time steps that are more than a threshold number of time steps after the particular time step in the sequence; and training, through reinforcement learning, the neural network system using at least the modified rewards for the particular time steps.

公开(公告)号：US20230178076A1

公开(公告)日：2023-06-08

申请号：US18077194

申请日：2022-12-07

Applicant: DeepMind Technologies Limited

Inventor： Joshua Simon Abramson ,  Arun Ahuja ,  Federico Javier Carnevale ,  Petko Ivanov Georgiev ,  Chia-Chun Hung ,  Timothy Paul Lillicrap ,  Alistair Michael Muldal ,  Adam Anthony Santoro ,  Tamara Louise von Glehn ,  Jessica Paige Landon ,  Gregory Duncan Wayne ,  Chen Yan ,  Rui Zhu

IPC: G10L15/22 ,  G10L15/16 ,  G10L13/02 ,  G06V10/82 ,  G06V20/50 ,  G06F40/284 ,  G06F40/40 ,  G06V10/774 ,  G10L15/06

CPC classification number: G10L15/22 ,  G10L15/16 ,  G10L13/02 ,  G06V10/82 ,  G06V20/50 ,  G06F40/284 ,  G06F40/40 ,  G06V10/774 ,  G10L15/063 ,  G10L2015/223

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for controlling agents. In particular, an interactive agent can be controlled based on multi-modal inputs that include both an observation image and a natural language text sequence.

公开(公告)号：US20200117956A1

公开(公告)日：2020-04-16

申请号：US16601324

申请日：2019-10-14

Applicant: DeepMind Technologies Limited

Inventor： Gregory Duncan Wayne ,  Timothy Paul Lillicrap ,  Chia-Chun Hung ,  Joshua Simon Abramson

IPC: G06K9/62 ,  G06N3/08 ,  G06F11/30

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network system used to control an agent interacting with an environment to perform a specified task. One of the methods includes causing the agent to perform a task episode in which the agent attempts to perform the specified task; for each of one or more particular time steps in the sequence: generating a modified reward for the particular time step from (i) the actual reward at the time step and (ii) value predictions at one or more time steps that are more than a threshold number of time steps after the particular time step in the sequence; and training, through reinforcement learning, the neural network system using at least the modified rewards for the particular time steps.