FlowGNN: A Dataflow Architecture for Real-Time Workload-Agnostic Graph Neural Network Inference.
Rishov SarkarStefan Abi-KaramYuqi HeLakshmi SathideviCong HaoPublished in: HPCA (2023)
Keyphrases
- real time
- neural network
- inference engine
- network architecture
- data flow
- vlsi architecture
- graph transformation
- graph structure
- directed graph
- database machine
- event driven
- design methodology
- random walk
- control system
- multi layer
- low cost
- bayesian networks
- pattern recognition
- parallel computing
- back propagation
- management system
- response time
- dedicated hardware
- complex event processing
- neural architecture
- structured data
- inference process
- software architecture
- artificial neural networks
- graph matching
- directed acyclic graph
- graph theory
- neural network model
- fuzzy neural network
- graph representation
- probabilistic model
- probabilistic inference
- weighted graph
- knn
- dynamic reconfiguration
- clique tree
- bipartite graph