DeepMind Technologies Limited patent applications on March 20th, 2025
Patent Applications by DeepMind Technologies Limited on March 20th, 2025
DeepMind Technologies Limited: 5 patent applications
DeepMind Technologies Limited has applied for patents in the areas of G05B13/02 (1), G06F40/58 (1), G06N3/09 (1), G06N3/092 (1), G06F3/033 (1) G05B13/027 (1), G06F3/033 (1), G06F30/27 (1), G06F30/327 (1), G06N3/045 (1)
With keywords such as: computer, control, network, neural, action, particular, input, training, including, and selection in patent application abstracts.
Patent Applications by DeepMind Technologies Limited
Inventor(s): Arun Ahuja of London GB for deepmind technologies limited, Robert David Fergus of New York NY US for deepmind technologies limited, Ishita Dasgupta of Brooklyn NY US for deepmind technologies limited, Kavya Venkata Kota Sai Kopparapu of New York NY US for deepmind technologies limited
IPC Code(s): G05B13/02, G06F40/58, G06N3/09, G06N3/092
CPC Code(s): G05B13/027
Abstract: methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training a high-level controller neural network for controlling an agent. in particular, the high-level controller neural network generates natural language commands that can be provided as input to a low-level controller neural network, which generates control outputs that can be used to control the agent.
Inventor(s): Peter Conway Humphreys of London GB for deepmind technologies limited, Timothy Paul Lillicrap of London GB for deepmind technologies limited, Tobias Markus Pohlen of London GB for deepmind technologies limited, Adam Anthony Santoro of London GB for deepmind technologies limited
IPC Code(s): G06F3/033, G06F3/023, G06F40/284
CPC Code(s): G06F3/033
Abstract: a computer-implemented method for controlling a particular computer to execute a task is described. the method includes receiving a control input comprising a visual input, the visual input including one or more screen frames of a computer display that represent at least a current state of the particular computer; processing the control input using a neural network to generate one or more control outputs that are used to control the particular computer to execute the task, in which the one or more control outputs include an action type output that specifies at least one of a pointing device action or a keyboard action to be performed to control the particular computer; determining one or more actions from the one or more control outputs; and executing the one or more actions to control the particular computer.
Inventor(s): Ian Michael Gemp of London GB for deepmind technologies limited, Yoram Bachrach of London GB for deepmind technologies limited
IPC Code(s): G06F30/27
CPC Code(s): G06F30/27
Abstract: methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating suggested communications during a multi-agent interaction using a language model neural network.
Inventor(s): Zhigang Pan of Austin TX US for deepmind technologies limited, Hanchen Ye of Sunnyvale CA US for deepmind technologies limited, Xiaoqing Xu of Mountain View CA US for deepmind technologies limited, Christopher Daniel Leary of Sunnyvale CA US for deepmind technologies limited
IPC Code(s): G06F30/327
CPC Code(s): G06F30/327
Abstract: the technology employs an iterative system of difference constraints (isdc) approach that leverages low-level feedback from downstream tools to iteratively refine scheduling with respect to circuit design high-level synthesis. in each iteration, a number of subgraphs are extracted from an original computation graph and passed to selected downstream tools, e.g., for logic synthesis, placement and/or routing. the downstream tools' compilation results are extracted and fed back to a scheduler. with the feedback, the scheduler recalculates delay estimation between each pair of nodes in the original computation graph and prunes redundant scheduling constraints. as a result, the explorable design space is enlarged in the next iteration, leading to refined scheduling results. this feedback-guided approach is compatible with versatile design constraints and objectives, such as minimizing register usage given a targeted clock period, minimizing the clock period given a constrained area budget, etc., to provide improvements to the system operation.
Inventor(s): Ziyu Wang of MARKHAM CA for deepmind technologies limited, Nicolas Manfred Otto Heess of London GB for deepmind technologies limited, Victor Constant Bapst of London GB for deepmind technologies limited
IPC Code(s): G06N3/045, G06N3/006, G06N3/047, G06N3/084, G06N3/088
CPC Code(s): G06N3/045
Abstract: methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an action selection neural network. one of the methods includes maintaining a replay memory that stores trajectories generated as a result of interaction of an agent with an environment; and training an action selection neural network having policy parameters on the trajectories in the replay memory, wherein training the action selection neural network comprises: sampling a trajectory from the replay memory; and adjusting current values of the policy parameters by training the action selection neural network on the trajectory using an off-policy actor critic reinforcement learning technique.
DeepMind Technologies Limited patent applications on March 20th, 2025
