Speedup Critical Stage of Machine Learning with Batch Scheduling in GPU
Abstract
As a superior data analysis method, Machine Learning suffers the bottleneck from limited computing capability for many years. With the advent of numerous parallel computing hardwares, modern GPU is becoming a promising carrier for the tasks of Machine Learning. In this paper, we propose an efficient GPU execution framework to speedup the forward propagation process of convolution neural network. By extending the convolution unrolling method to fit this batch mode, we get a significant increase of throughput but very little overhead.
Domains
Computer Science [cs]
Origin : Files produced by the author(s)
