CIMGN: An Energy-efficient All-digital Compute-in-memory Graph Neural Network Processor.
Yipeng WangMengtian YangShanshan XieMeizhi WangJaydeep P. KulkarniPublished in: ESSCIRC (2023)
Keyphrases
- neural network
- auto associative
- intel xeon
- wireless sensor networks
- memory management
- graph theory
- random walk
- pattern recognition
- memory hierarchy
- memory subsystem
- energy efficient
- associative memory
- back propagation
- computing power
- graph representation
- neural network model
- weighted graph
- genetic algorithm
- memory space
- graph model
- graph structure
- parallel processing
- directed graph
- high speed
- structured data
- energy efficiency
- sensor networks
- main memory
- random access memory
- self organizing maps
- graph search
- instruction set
- fault diagnosis
- memory access
- artificial neural networks
- neural network is trained
- multithreading
- processing elements
- distributed memory
- random access
- memory requirements
- graph mining
- directed acyclic graph