18589092. COMPRESSING A NEURAL NETWORK simplified abstract (Imagination Technologies Limited)

From WikiPatents
Jump to navigation Jump to search

COMPRESSING A NEURAL NETWORK

Organization Name

Imagination Technologies Limited

Inventor(s)

Gunduz Vehbi Demirci of Hertfordshire (GB)

Cagatay Dikici of Hertfordshire (GB)

COMPRESSING A NEURAL NETWORK - A simplified explanation of the abstract

This abstract first appeared for US patent application 18589092 titled 'COMPRESSING A NEURAL NETWORK

The abstract describes a method for compressing a neural network by rearranging the elements of a matrix representing a layer of the network to create sub-matrices with a higher density of non-zero values.

  • The method involves receiving a neural network and determining a matrix of coefficients for a layer.
  • The matrix is rearranged to group non-zero values into sub-matrices with higher density.
  • The compressed neural network includes a compressed layer that performs operations based on the sub-matrices.

Potential Applications: - Efficient storage and transmission of neural networks. - Faster inference and reduced computational requirements. - Improved performance in resource-constrained environments.

Problems Solved: - Addressing the challenge of large neural network sizes. - Enhancing the scalability and practicality of neural network deployment. - Optimizing neural network operations for better efficiency.

Benefits: - Reduced memory and processing requirements. - Faster execution of neural network tasks. - Enhanced performance in edge computing and IoT devices.

Commercial Applications: Title: "Efficient Neural Network Compression for Enhanced Performance" This technology can be applied in industries such as: - Mobile devices for faster AI applications. - Cloud computing for optimized resource utilization. - Autonomous vehicles for real-time decision-making.

Questions about Neural Network Compression: 1. How does rearranging the matrix elements improve neural network compression? 2. What are the implications of using sub-matrices for compressed neural network operations?

Frequently Updated Research: Stay updated on advancements in neural network compression techniques and their impact on AI applications.


Original Abstract Submitted

A computer implemented method of compressing a neural network, the method comprising: receiving a neural network; determining a matrix representative of a set of coefficients of a layer of the received neural network, the layer being arranged to perform an operation, the matrix comprising a plurality of elements representative of non-zero values and a plurality of elements representative of zero values; rearranging the rows and/or columns of the matrix so as to gather the plurality of elements representative of non-zero values of the matrix into one or more sub-matrices, the one or more sub-matrices having a greater number of elements representative of non-zero values per total number of elements of the one or more sub-matrices than the number of elements representative of non-zero values per total number of elements of the matrix; and outputting a compressed neural network comprising a compressed layer arranged to perform a compressed operation in dependence on the one or more sub-matrices.

Retrieved from "http://wikipatents.org/index.php?title=18589092._COMPRESSING_A_NEURAL_NETWORK_simplified_abstract_(Imagination_Technologies_Limited)&oldid=244259"