Deepmind technologies limited (20250103856). LOCAL CROSS-ATTENTION OPERATIONS IN NEURAL NETWORKS

LOCAL CROSS-ATTENTION OPERATIONS IN NEURAL NETWORKS

Organization Name

deepmind technologies limited

Inventor(s)

Joao Carreira of St. Albans GB

Andrew Coulter Jaegle of London GB

Skanda Kumar Koppula of London GB

Daniel Zoran of London GB

Adrià Recasens Continente of London GB

Catalin-Dumitru Ionescu of London GB

[[:Category:Olivier Jean H�naff of London GB|Olivier Jean H�naff of London GB]][[Category:Olivier Jean H�naff of London GB]]

Evan Gerard Shelhamer of London GB

Relja Arandjelovic of Munich DE

Matthew Botvinick of Philadelphia PA US

Oriol Vinyals of London GB

Karen Simonyan of London GB

Andrew Zisserman of London GB

LOCAL CROSS-ATTENTION OPERATIONS IN NEURAL NETWORKS

This abstract first appeared for US patent application 20250103856 titled 'LOCAL CROSS-ATTENTION OPERATIONS IN NEURAL NETWORKS

Original Abstract Submitted

methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for using a neural network to generate a network output that characterizes an entity. in one aspect, a method includes: obtaining a representation of the entity as a set of data element embeddings, obtaining a set of latent embeddings, and processing: (i) the set of data element embeddings, and (ii) the set of latent embeddings, using the neural network to generate the network output. the neural network includes a sequence of neural network blocks including: (i) one or more local cross-attention blocks, and (ii) an output block. each local cross-attention block partitions the set of latent embeddings and the set of data element embeddings into proper subsets, and updates each proper subset of the set of latent embeddings using attention over only the corresponding proper subset of the set of data element embeddings.