Online Learning for DNN Training: A Stochastic Block Adaptive Gradient Algorithm

<div>Training accuracy with running time for ResNet-34 and DenseNet-121 on CIFAR-10: a comparative summary. (a) Training accuracy for ResNet-34. (b) Training accuracy for DenseNet-121.</div>

Computational Intelligence and Neuroscience

fig3

Figure 3

Figure 3: Online Learning for DNN Training: A Stochastic Block Adaptive Gradient Algorithm 