Revisiting Convergence of AdaGrad with Relaxed Assumptions
This work provides improved theoretical guarantees for a widely used optimization algorithm, addressing realistic noise assumptions in machine learning applications, though it is incremental as it builds on existing AdaGrad analysis.
The authors revisited the convergence of AdaGrad with momentum for non-convex smooth optimization under a general noise model, achieving a probabilistic rate of O~(1/√T) that can accelerate to O~(1/T) under low noise conditions, matching known lower bounds up to logarithmic terms.
In this study, we revisit the convergence of AdaGrad with momentum (covering AdaGrad as a special case) on non-convex smooth optimization problems. We consider a general noise model where the noise magnitude is controlled by the function value gap together with the gradient magnitude. This model encompasses a broad range of noises including bounded noise, sub-Gaussian noise, affine variance noise and the expected smoothness, and it has been shown to be more realistic in many practical applications. Our analysis yields a probabilistic convergence rate which, under the general noise, could reach at (\tilde{\mathcal{O}}(1/\sqrt{T})). This rate does not rely on prior knowledge of problem-parameters and could accelerate to (\tilde{\mathcal{O}}(1/T)) where (T) denotes the total number iterations, when the noise parameters related to the function value gap and noise level are sufficiently small. The convergence rate thus matches the lower rate for stochastic first-order methods over non-convex smooth landscape up to logarithm terms [Arjevani et al., 2023]. We further derive a convergence bound for AdaGrad with mometum, considering the generalized smoothness where the local smoothness is controlled by a first-order function of the gradient norm.