Deepmind Technologies Limited

Facebook Technologies, LLC

Gdm Holding LLC

Joshua Merel

Vu Pham

Alexandre Galashov

Leonard Hasenclever

Patrick Kaifosh

Nicolas Manfred Otto Heess

Gregory Duncan Wayne

Raia Thais Hadsell

Thomas Reardon

Martin Riedmiller

Ziyu Wang

Peter William Battaglia

Timothy Machado

Erik Schomburg

Steven Brian Demers

Alvaro Sanchez

Saran Tunyasuvunakool

János Kramár

Jost Tobias Springenberg

Joao Ferdinando Freitas

Scott Ellison Reed

Yuke Zhu

Nicholas Manfred Otto Heess

A graph neural network system for processing data representing a physical system is disclosed. The graph neural network system receives physical system data and action data, where the physical system includes bodies coupled by joints. The graph neural network system processes an input graph to provide an output graph, the input and output graphs each having a graph structure including nodes and edges corresponding, respectively, to the bodies and joints of the physical system. The graph processing neural network provides the inferred static property or dynamic state of the physical system. The graph processing neural network is differentiable and may be used for control and/or reinforcement learning.

Graph neural networks representing physical systems

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.

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

Methods and apparatus for adapting a control mapping associating sensor signals with control signals for controlling an operation of a device. The method comprises obtaining first state information for an operation of the device, providing the first state information as input to an intention model associated with an operation of the device and obtaining corresponding first intention model output, providing a plurality of neuromuscular signals recorded from a user and/or signals derived from the neuromuscular signals as inputs to a first control mapping and obtaining corresponding first control mapping output, and updating the first control mapping using the inputs provided to the first control mapping and the first intention model output to obtain a second control mapping.

Adaptive system for deriving control signals from measurements of neuromuscular activity

Growing community of inventors

London, United Kingdom

Joshua Merel