3 papers
cs.CV2018
A predictor-corrector method for the training of deep neural networks
Yatin Saraiya
The training of deep neural nets is expensive. We present a predictor- corrector method for the training of deep neural nets. It alternates a predictor pass with a corrector pass u…
cs.CV2018
Using accumulation to optimize deep residual neural nets
Yatin Saraiya
Residual Neural Networks [1] won first place in all five main tracks of the ImageNet and COCO 2015 competitions. This kind of network involves the creation of pluggable modules suc…
cs.LG2018
Leapfrogging for parallelism in deep neural networks
Yatin Saraiya
We present a technique, which we term leapfrogging, to parallelize back- propagation in deep neural networks. We show that this technique yields a savings of of a dominant…