Noise Is Not the Main Factor Behind the Gap Between SGD and Adam on Transformers, but Sign Descent Might Be

• URL: http://arxiv.org/abs/2304.13960v1
• Date: Thu, 27 Apr 2023 05:41:13 GMT
• Title: Noise Is Not the Main Factor Behind the Gap Between SGD and Adam on Transformers, but Sign Descent Might Be
• Authors: Frederik Kunstner, Jacques Chen, Jonathan Wilder Lavington, Mark Schmidt
• Abstract summary: We show that the behavior of Adam with large batches is similar to sign descent with momentum. We present evidence thatity and heavy-tailed noise are not major factors in the performance gap between SGD and Adam.
• Score: 16.170888329408353
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The success of the Adam optimizer on a wide array of architectures has made it the default in settings where stochastic gradient descent (SGD) performs poorly. However, our theoretical understanding of this discrepancy is lagging, preventing the development of significant improvements on either algorithm. Recent work advances the hypothesis that Adam and other heuristics like gradient clipping outperform SGD on language tasks because the distribution of the error induced by sampling has heavy tails. This suggests that Adam outperform SGD because it uses a more robust gradient estimate. We evaluate this hypothesis by varying the batch size, up to the entire dataset, to control for stochasticity. We present evidence that stochasticity and heavy-tailed noise are not major factors in the performance gap between SGD and Adam. Rather, Adam performs better as the batch size increases, while SGD is less effective at taking advantage of the reduction in noise. This raises the question as to why Adam outperforms SGD in the full-batch setting. Through further investigation of simpler variants of SGD, we find that the behavior of Adam with large batches is similar to sign descent with momentum.

Related papers

• Deconstructing What Makes a Good Optimizer for Language Models [7.9224468703944115]
  We compare several optimization algorithms, including SGD, Adafactor, Adam, and Lion, in the context of autoregressive language modeling. Our findings indicate that, except for SGD, these algorithms all perform comparably both in their optimal performance.
  arXiv  Detail & Related papers  (2024-07-10T18:11:40Z)
• On Convergence of Adam for Stochastic Optimization under Relaxed Assumptions [4.9495085874952895]
  Adaptive Momentum Estimation (Adam) algorithm is highly effective in various deep learning tasks. We show that Adam can find a stationary point variance with a rate in high iterations under this general noise model.
  arXiv  Detail & Related papers  (2024-02-06T13:19:26Z)
• Provable Adaptivity of Adam under Non-uniform Smoothness [79.25087082434975]
  Adam is widely adopted in practical applications due to its fast convergence. Existing convergence analyses for Adam rely on the bounded smoothness assumption. This paper studies the convergence of randomly reshuffled Adam with diminishing learning rate.
  arXiv  Detail & Related papers  (2022-08-21T14:57:47Z)
• Understanding AdamW through Proximal Methods and Scale-Freeness [57.47324825501137]
  Adam is a generalization of the $ell$ regularizer Adam-$ell$. AdamW decouples the gradient of Adam-$ell$ from the update rule of Adam-$ell$. We show that AdamW exhibits an advantage over Adam-$ell$ and the degree to which we expect the gradients of the network to exhibit multiple scales.
  arXiv  Detail & Related papers  (2022-01-31T21:00:55Z)
• Why Does Multi-Epoch Training Help? [62.946840431501855]
  Empirically, it has been observed that taking more one pass over training data (multi-pass SGD) has much better excess risk bound performance than SGD only taking one pass over training data (one-pass SGD) In this paper, we provide some theoretical evidences for explaining why multiple passes over the training data can help improve performance under certain circumstances.
  arXiv  Detail & Related papers  (2021-05-13T00:52:25Z)
• Correcting Momentum with Second-order Information [50.992629498861724]
  We develop a new algorithm for non-critical optimization that finds an $O(epsilon)$epsilon point in the optimal product. We validate our results on a variety of large-scale deep learning benchmarks and architectures.
  arXiv  Detail & Related papers  (2021-03-04T19:01:20Z)
• Adam$^+$: A Stochastic Method with Adaptive Variance Reduction [56.051001950733315]
  Adam is a widely used optimization method for deep learning applications. We propose a new method named Adam$+$ (pronounced as Adam-plus) Our empirical studies on various deep learning tasks, including image classification, language modeling, and automatic speech recognition, demonstrate that Adam$+$ significantly outperforms Adam.
  arXiv  Detail & Related papers  (2020-11-24T09:28:53Z)
• AdaSGD: Bridging the gap between SGD and Adam [14.886598905466604]
  We identify potential differences in performance between SGD and Adam. We demonstrate how AdaSGD combines the benefits both SGD Adam and SGD non- descent.
  arXiv  Detail & Related papers  (2020-06-30T05:44:19Z)
• AdamP: Slowing Down the Slowdown for Momentum Optimizers on Scale-invariant Weights [53.8489656709356]
  Normalization techniques are a boon for modern deep learning. It is often overlooked, however, that the additional introduction of momentum results in a rapid reduction in effective step sizes for scale-invariant weights. In this paper, we verify that the widely-adopted combination of the two ingredients lead to the premature decay of effective step sizes and sub-optimal model performances.
  arXiv  Detail & Related papers  (2020-06-15T08:35:15Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.