View source for 18814438. ASYNCHRONOUS NEURAL NETWORK TRAINING (Microsoft Technology Licensing, LLC)

=ASYNCHRONOUS NEURAL NETWORK TRAINING=


==Organization Name==

[[:Category:Microsoft Technology Licensing, LLC|Microsoft Technology Licensing, LLC]]


[[Category:Microsoft Technology Licensing, LLC]]

==Inventor(s)==

[[:Category:Ryota Tomioka of Cambridge (GB)|Ryota Tomioka of Cambridge (GB)]][[Category:Ryota Tomioka of Cambridge (GB)]]

[[:Category:Matthew Alastair Johnson of Cambridge (GB)|Matthew Alastair Johnson of Cambridge (GB)]][[Category:Matthew Alastair Johnson of Cambridge (GB)]]

[[:Category:Daniel Stefan Tarlow of Cambridge (GB)|Daniel Stefan Tarlow of Cambridge (GB)]][[Category:Daniel Stefan Tarlow of Cambridge (GB)]]

[[:Category:Samuel Alexander Webster of Cambridge (GB)|Samuel Alexander Webster of Cambridge (GB)]][[Category:Samuel Alexander Webster of Cambridge (GB)]]

[[:Category:Dimitrios Vytiniotis of Cambridge (GB)|Dimitrios Vytiniotis of Cambridge (GB)]][[Category:Dimitrios Vytiniotis of Cambridge (GB)]]

[[:Category:Alexander Lloyd Gaunt of Cambridge (GB)|Alexander Lloyd Gaunt of Cambridge (GB)]][[Category:Alexander Lloyd Gaunt of Cambridge (GB)]]

[[:Category:Maik Riechert of Cambridge (GB)|Maik Riechert of Cambridge (GB)]][[Category:Maik Riechert of Cambridge (GB)]]

==ASYNCHRONOUS NEURAL NETWORK TRAINING==


This abstract first appeared for US patent application 18814438 titled 'ASYNCHRONOUS NEURAL NETWORK TRAINING




==Original Abstract Submitted==

A neural network training apparatus is described which has a network of worker nodes each having a memory storing a subgraph of a neural network to be trained. The apparatus has a control node connected to the network of worker nodes. The control node is configured to send training data instances into the network to trigger parallelized message passing operations which implement a training algorithm which trains the neural network. At least some of the message passing operations asynchronously update parameters of individual subgraphs of the neural network at the individual worker nodes.



[[Category:G06N3/08]]
[[Category:G06N3/04]]
[[Category:G06N3/063]]
[[Category:CPC_G06N3/08]]