Jethro's Braindump

LARS Optimizer

Layer-wise Adaptive Rate Scaling (LARS) is a §nn_optimizer. The technique allows §large_batch_training without significant decrease in accuracy (You et al., 2017). One of the secondary goals is §fast_nn_training.



You, Y., Gitman, I., & Ginsburg, B., Large batch training of convolutional networks, CoRR, (), (2017).

Icon by Laymik from The Noun Project. Website built with ♥ with Org-mode, Hugo, and Netlify.