Oct 29, 2017 · In this paper, we develop control variate based algorithms which allow sampling an arbitrarily small neighbor size.
We present a novel control variate based algorithm that uti- lizes historical activations to reduce the estimator variance. 3.1. Control Variate Based Estimator.
Nov 14, 2017 · A control variate based stochastic training algorithm for graph convolutional networks that the receptive field can be only two neighbors per node.
Code for the paper Stochastic Training for Graph Convolutional Networks. The implementation is based on Thomas Kipf's implementation for graph convolutional ...
The following lemma bounds the approximation error of activations in a multi- layer GCN with CV. Intuitively, there is a sequence of slow-changing model.
Control variate based algorithms which allow sampling an arbitrarily small neighbor size are developed and a new theoretical guarantee for these algorithms ...
We empirically test on six graph datasets, and show that our techniques significantly reduce the bias and variance of the gradient from NS with the same ...
Jul 22, 2020 · Jianfei Chen, Jun Zhu, Le Song: Stochastic Training of Graph Convolutional Networks with Variance Reduction. ICML 2018: 941-949.
A preprocessing strategy and two control variate based algorithms to further reduce the receptive field size of graph convolutional networks and are ...
Oct 28, 2017 · StoGCN is a control variate based algorithm which allow sampling an arbitrarily small neighbor size. Presents new theoretical guarantee for the algorithms to ...