Learn Locally, Correct Globally: A Distributed Algorithm for Training Graph Neural Networks.
Morteza RamezaniWeilin CongMehrdad MahdaviMahmut T. KandemirAnand SivasubramaniamPublished in: CoRR (2021)
Keyphrases
- detection algorithm
- neural network
- learning algorithm
- training algorithm
- training process
- worst case
- computational complexity
- preprocessing
- tree structure
- globally consistent
- graph partitioning
- cost function
- np hard
- dynamic programming
- graph model
- learning rules
- graph structure
- matching algorithm
- artificial neural networks
- k means
- segmentation algorithm
- minimum spanning tree
- decision trees
- feedforward neural networks
- feed forward neural networks
- maximum flow
- backpropagation algorithm
- similarity measure
- training phase
- globally optimal
- spanning tree
- reinforcement learning
- optimal solution
- weighted graph
- convergence rate
- random walk
- expectation maximization
- particle swarm optimization