Theory

Bypassing Stationary Points in Training Deep Learning Models
Gradient-descent-based optimizers are prone to slowdowns in training deep learning models, as stationary points are ubiquitous in the …
Cite DOI URL