A Theoretical and Empirical Study on the Convergence of Adam with an "Exact" Constant Step Size in Non-Convex Settings
This work addresses a theoretical gap for researchers and practitioners using Adam in neural network training, though it is incremental as it builds on existing adaptive gradient methods.
The paper tackles the problem of Adam's convergence in non-convex optimization by theoretically analyzing and empirically validating a constant step size, showing it achieves asymptotic gradient convergence to zero and outperforms state-of-the-art schedulers in reducing gradient norms.
In neural network training, RMSProp and Adam remain widely favoured optimisation algorithms. One of the keys to their performance lies in selecting the correct step size, which can significantly influence their effectiveness. Additionally, questions about their theoretical convergence properties continue to be a subject of interest. In this paper, we theoretically analyse a constant step size version of Adam in the non-convex setting and discuss why it is important for the convergence of Adam to use a fixed step size. This work demonstrates the derivation and effective implementation of a constant step size for Adam, offering insights into its performance and efficiency in non convex optimisation scenarios. (i) First, we provide proof that these adaptive gradient algorithms are guaranteed to reach criticality for smooth non-convex objectives with constant step size, and we give bounds on the running time. Both deterministic and stochastic versions of Adam are analysed in this paper. We show sufficient conditions for the derived constant step size to achieve asymptotic convergence of the gradients to zero with minimal assumptions. Next, (ii) we design experiments to empirically study Adam's convergence with our proposed constant step size against stateof the art step size schedulers on classification tasks. Lastly, (iii) we also demonstrate that our derived constant step size has better abilities in reducing the gradient norms, and empirically, we show that despite the accumulation of a few past gradients, the key driver for convergence in Adam is the non-increasing step sizes.