Timezone: »

 
Poster
High-probability Bounds for Non-Convex Stochastic Optimization with Heavy Tails
Ashok Cutkosky · Harsh Mehta

Thu Dec 09 04:30 PM -- 06:00 PM (PST) @
in Poster Session 7 »
We consider non-convex stochastic optimization using first-order algorithms for which the gradient estimates may have heavy tails. We show that a combination of gradient clipping, momentum, and normalized gradient descent yields convergence to critical points in high-probability with best-known rates for smooth losses when the gradients only have bounded $\mathfrak{p}$th moments for some $\mathfrak{p}\in(1,2]$. We then consider the case of second-order smooth losses, which to our knowledge have not been studied in this setting, and again obtain high-probability bounds for any $\mathfrak{p}$. Moreover, our results hold for arbitrary smooth norms, in contrast to the typical SGD analysis which requires a Hilbert space norm. Further, we show that after a suitable "burn-in" period, the objective value will monotonically decrease for every iteration until a critical point is identified, which provides intuition behind the popular practice of learning rate "warm-up'' and also yields a last-iterate guarantee.
Keywords: Deep Learning · Optimization

Author Information

Ashok Cutkosky (Boston University)
Harsh Mehta (Google Research)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors