Scratching Visual Transformer's Back with Uniform Attention

• URL: http://arxiv.org/abs/2210.08457v2
• Date: Thu, 26 Dec 2024 02:16:35 GMT
• Title: Scratching Visual Transformer's Back with Uniform Attention
• Abstract summary: We study the role of the density of the attention in ViT models.<n>We manually insert the uniform attention to each layer of ViT models to supply the much needed dense interactions.<n>We observe that the inclusion of CB reduces the degree of density in the original attention maps and increases both the capacity and generalizability of the ViT models.
• Score: 49.31658397277271
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The favorable performance of Vision Transformers (ViTs) is often attributed to the multi-head self-attention (MSA). The MSA enables global interactions at each layer of a ViT model, which is a contrasting feature against Convolutional Neural Networks (CNNs) that gradually increase the range of interaction across multiple layers. We study the role of the density of the attention. Our preliminary analyses suggest that the spatial interactions of attention maps are close to dense interactions rather than sparse ones. This is a curious phenomenon, as dense attention maps are harder for the model to learn due to steeper softmax gradients around them. We interpret this as a strong preference for ViT models to include dense interaction. We thus manually insert the uniform attention to each layer of ViT models to supply the much needed dense interactions. We call this method Context Broadcasting, CB. We observe that the inclusion of CB reduces the degree of density in the original attention maps and increases both the capacity and generalizability of the ViT models. CB incurs negligible costs: 1 line in your model code, no additional parameters, and minimal extra operations.

Related papers

• Look Less, Reason More: Block-wise Attention Skipping for Efficient Multimodal LLMs [89.7106332677868]
  We present Visual-Skip, a training-free inference paradigm that decouples spatial interaction from semantic evolution.<n>V-Skip effectively bypasses redundant vision attention to achieve block-wise sparsity, maintaining a 94.16% to 100.31% performance retention across diverse MLLMs.
  arXiv  Detail & Related papers  (2026-06-07T08:32:13Z)
• Architecture Decoupling Is Not All You Need For Unified Multimodal Model [64.19284951218098]
  We propose Attention Interaction Alignment (AIA) loss, which explicitly learns Task-Specific multimodal interaction patterns during training.<n>AIA not only refines cross-modal attention patterns, but also boosts both generation and understanding performance.
  arXiv  Detail & Related papers  (2025-11-27T17:55:25Z)
• Super Consistency of Neural Network Landscapes and Learning Rate Transfer [72.54450821671624]
  We study the landscape through the lens of the loss Hessian. We find that certain spectral properties under $mu$P are largely independent of the size of the network. We show that in the Neural Tangent Kernel (NTK) and other scaling regimes, the sharpness exhibits very different dynamics at different scales.
  arXiv  Detail & Related papers  (2024-02-27T12:28:01Z)
• Accelerating Vision Transformers Based on Heterogeneous Attention Patterns [89.86293867174324]
  Vision Transformers (ViTs) have attracted a lot of attention in the field of computer vision. We propose an integrated compression pipeline based on observed heterogeneous attention patterns across layers. Experimentally, the integrated compression pipeline of DGSSA and GLAD can accelerate up to 121% run-time throughput.
  arXiv  Detail & Related papers  (2023-10-11T17:09:19Z)
• A Close Look at Spatial Modeling: From Attention to Convolution [70.5571582194057]
  Vision Transformers have shown great promise recently for many vision tasks due to the insightful architecture design and attention mechanism. We generalize self-attention formulation to abstract a queryirrelevant global context directly and integrate the global context into convolutions. With less than 14M parameters, our FCViT-S12 outperforms related work ResT-Lite by 3.7% top1 accuracy on ImageNet-1K.
  arXiv  Detail & Related papers  (2022-12-23T19:13:43Z)
• Convolution-enhanced Evolving Attention Networks [41.684265133316096]
  Evolving Attention-enhanced Dilated Convolutional (EA-DC-) Transformer outperforms state-of-the-art models significantly. This is the first work that explicitly models the layer-wise evolution of attention maps.
  arXiv  Detail & Related papers  (2022-12-16T08:14:04Z)
• Vision Transformers provably learn spatial structure [34.61885883486938]
  Vision Transformers (ViTs) have achieved comparable or superior performance than Convolutional Neural Networks (CNNs) in computer vision. Yet, recent works have shown that while minimizing their training loss, ViTs specifically learn spatially localized patterns.
  arXiv  Detail & Related papers  (2022-10-13T19:53:56Z)
• On Improving Adversarial Transferability of Vision Transformers [97.17154635766578]
  Vision transformers (ViTs) process input images as sequences of patches via self-attention. We study the adversarial feature space of ViT models and their transferability. We introduce two novel strategies specific to the architecture of ViT models.
  arXiv  Detail & Related papers  (2021-06-08T08:20:38Z)
• DeepViT: Towards Deeper Vision Transformer [92.04063170357426]
  Vision transformers (ViTs) have been successfully applied in image classification tasks recently. We show that, unlike convolution neural networks (CNNs)that can be improved by stacking more convolutional layers, the performance of ViTs saturate fast when scaled to be deeper. We propose a simple yet effective method, named Re-attention, to re-generate the attention maps to increase their diversity.
  arXiv  Detail & Related papers  (2021-03-22T14:32:07Z)
• GTA: Global Temporal Attention for Video Action Understanding [51.476605514802806]
  We introduce Global Temporal Attention (AGT), which performs global temporal attention on top of spatial attention in a decoupled manner. Tests on 2D and 3D networks demonstrate that our approach consistently enhances temporal modeling and provides state-of-the-art performance on three video action recognition datasets.
  arXiv  Detail & Related papers  (2020-12-15T18:58:21Z)

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.