cheng fang
Rutgers univerisity
student-facilitator
HPC Experience
Design a decentralized algorithm of machine learning/optimization to solve the issue either because the local computational power is limited or data is not acquirable for the central server. Have theoretical proof of this algorithm to converge to the optimal point and run this algorithm (with MNIST AND CIFAR dataset) on amarel cluster (Rutgers computational facility) to experimentally show this decentralized algorithm could alleviate the problem mentioned above.