Pay Attention to Small Weights

• URL: http://arxiv.org/abs/2506.21374v1
• Date: Thu, 26 Jun 2025 15:22:55 GMT
• Title: Pay Attention to Small Weights
• Authors: Chao Zhou, Tom Jacobs, Advait Gadhikar, Rebekka Burkholz,
• Abstract summary: NanoADAM dynamically updates only the small-magnitude weights during finetuning.<n>It preserves large-magnitude weights, which are likely to encode critical features learned during pretraining.
• Score: 16.13381473792083
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Finetuning large pretrained neural networks is known to be resource-intensive, both in terms of memory and computational cost. To mitigate this, a common approach is to restrict training to a subset of the model parameters. By analyzing the relationship between gradients and weights during finetuning, we observe a notable pattern: large gradients are often associated with small-magnitude weights. This correlation is more pronounced in finetuning settings than in training from scratch. Motivated by this observation, we propose NANOADAM, which dynamically updates only the small-magnitude weights during finetuning and offers several practical advantages: first, this criterion is gradient-free -- the parameter subset can be determined without gradient computation; second, it preserves large-magnitude weights, which are likely to encode critical features learned during pretraining, thereby reducing the risk of catastrophic forgetting; thirdly, it permits the use of larger learning rates and consistently leads to better generalization performance in experiments. We demonstrate this for both NLP and vision tasks.

Related papers

• PACE: Marrying generalization in PArameter-efficient fine-tuning with Consistency rEgularization [35.922096876707975]
  PACE is a generalization of PArameter-efficient fine-tuning with Consistency rEgularization.<n>It implicitly regularizes gradients for enhanced generalization, but also implicitly aligns the fine-tuned and pre-trained models to retain knowledge.<n>It also improves LoRA in text classification (GLUE) and mathematical reasoning.
  arXiv  Detail & Related papers  (2024-09-25T17:56:00Z)
• Pre-Pruning and Gradient-Dropping Improve Differentially Private Image Classification [9.120531252536617]
  We introduce a new training paradigm that uses textitpre-pruning and textitgradient-dropping to reduce the parameter space and improve scalability. Our training paradigm introduces a tension between the rates of pre-pruning and gradient-dropping, privacy loss, and classification accuracy.
  arXiv  Detail & Related papers  (2023-06-19T14:35:28Z)
• SGD with Large Step Sizes Learns Sparse Features [22.959258640051342]
  We showcase important features of the dynamics of the Gradient Descent (SGD) in the training of neural networks. We show that the longer large step sizes keep SGD high in the loss landscape, the better the implicit regularization can operate and find sparse representations.
  arXiv  Detail & Related papers  (2022-10-11T11:00:04Z)
• Scaling Forward Gradient With Local Losses [117.22685584919756]
  Forward learning is a biologically plausible alternative to backprop for learning deep neural networks. We show that it is possible to substantially reduce the variance of the forward gradient by applying perturbations to activations rather than weights. Our approach matches backprop on MNIST and CIFAR-10 and significantly outperforms previously proposed backprop-free algorithms on ImageNet.
  arXiv  Detail & Related papers  (2022-10-07T03:52:27Z)
• Slimmable Networks for Contrastive Self-supervised Learning [69.9454691873866]
  Self-supervised learning makes significant progress in pre-training large models, but struggles with small models. We introduce another one-stage solution to obtain pre-trained small models without the need for extra teachers. A slimmable network consists of a full network and several weight-sharing sub-networks, which can be pre-trained once to obtain various networks.
  arXiv  Detail & Related papers  (2022-09-30T15:15:05Z)
• Training Thinner and Deeper Neural Networks: Jumpstart Regularization [2.8348950186890467]
  We use regularization to prevent neurons from dying or becoming linear. In comparison to conventional training, we obtain neural networks that are thinner, deeper, and - most importantly - more parameter-efficient.
  arXiv  Detail & Related papers  (2022-01-30T12:11:24Z)
• Powerpropagation: A sparsity inducing weight reparameterisation [65.85142037667065]
  We introduce Powerpropagation, a new weight- parameterisation for neural networks that leads to inherently sparse models. Models trained in this manner exhibit similar performance, but have a distribution with markedly higher density at zero, allowing more parameters to be pruned safely. Here, we combine Powerpropagation with a traditional weight-pruning technique as well as recent state-of-the-art sparse-to-sparse algorithms, showing superior performance on the ImageNet benchmark.
  arXiv  Detail & Related papers  (2021-10-01T10:03:57Z)
• Deep learning: a statistical viewpoint [120.94133818355645]
  Deep learning has revealed some major surprises from a theoretical perspective. In particular, simple gradient methods easily find near-perfect solutions to non-optimal training problems. We conjecture that specific principles underlie these phenomena.
  arXiv  Detail & Related papers  (2021-03-16T16:26:36Z)
• Neural networks with late-phase weights [66.72777753269658]
  We show that the solutions found by SGD can be further improved by ensembling a subset of the weights in late stages of learning. At the end of learning, we obtain back a single model by taking a spatial average in weight space.
  arXiv  Detail & Related papers  (2020-07-25T13:23:37Z)
• Extrapolation for Large-batch Training in Deep Learning [72.61259487233214]
  We show that a host of variations can be covered in a unified framework that we propose. We prove the convergence of this novel scheme and rigorously evaluate its empirical performance on ResNet, LSTM, and Transformer.
  arXiv  Detail & Related papers  (2020-06-10T08:22:41Z)

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.