A Computational-Graph Partitioning Method for Training Memory-Constrained DNNs.
Fareed QararyahMohamed WahibDoga DikbayirMehmet Esat BelviranliDidem UnatPublished in: CoRR (2020)
Keyphrases
- graph structure
- directed graph
- computational power
- training set
- training examples
- structured data
- graph theory
- connected components
- random walk
- memory requirements
- graph theoretic
- training phase
- graph model
- memory usage
- computing power
- graph representation
- memory space
- memory size
- bipartite graph
- training algorithm
- clustering algorithm
- pattern mining
- test set
- supervised learning
- objective function