Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factors

• URL: http://arxiv.org/abs/2103.15949v2
• Date: Tue, 4 Apr 2023 06:43:19 GMT
• Title: Transformer visualization via dictionary learning: contextualized embedding as a linear superposition of transformer factors
• Authors: Zeyu Yun, Yubei Chen, Bruno A Olshausen, Yann LeCun
• Abstract summary: We propose to use dictionary learning to open up "black boxes" as linear superpositions of transformer factors. Through visualization, we demonstrate the hierarchical semantic structures captured by the transformer factors. We hope this visualization tool can bring further knowledge and a better understanding of how transformer networks work.
• Score: 15.348047288817478
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Transformer networks have revolutionized NLP representation learning since they were introduced. Though a great effort has been made to explain the representation in transformers, it is widely recognized that our understanding is not sufficient. One important reason is that there lack enough visualization tools for detailed analysis. In this paper, we propose to use dictionary learning to open up these "black boxes" as linear superpositions of transformer factors. Through visualization, we demonstrate the hierarchical semantic structures captured by the transformer factors, e.g., word-level polysemy disambiguation, sentence-level pattern formation, and long-range dependency. While some of these patterns confirm the conventional prior linguistic knowledge, the rest are relatively unexpected, which may provide new insights. We hope this visualization tool can bring further knowledge and a better understanding of how transformer networks work. The code is available at https://github.com/zeyuyun1/TransformerVis

Related papers

• Extracting Finite State Machines from Transformers [0.3069335774032178]
  We investigate the trainability of transformers trained on regular languages from a mechanistic interpretability perspective. We empirically find tighter lower bounds on the trainability of transformers, when a finite number of symbols determine the state. Our mechanistic insight allows us to characterise the regular languages a one-layer transformer can learn with good length generalisation.
  arXiv  Detail & Related papers  (2024-10-08T13:43:50Z)
• Transformer Explainer: Interactive Learning of Text-Generative Models [65.91049787390692]
  Transformer Explainer is an interactive visualization tool designed for non-experts to learn about Transformers through the GPT-2 model. It runs a live GPT-2 instance locally in the user's browser, empowering users to experiment with their own input and observe in real-time how the internal components and parameters of the Transformer work together.
  arXiv  Detail & Related papers  (2024-08-08T17:49:07Z)
• Why transformers are obviously good models of language [19.214362677861025]
  Transformers are neural networks that process language automatically with more success than alternatives. I highlight direct connections between the transformer architecture and certain theoretical perspectives on language.
  arXiv  Detail & Related papers  (2024-08-07T15:52:46Z)
• Transformers are Expressive, But Are They Expressive Enough for Regression? [38.369337945109855]
  We show that Transformers struggle to reliably approximate smooth functions, relying on piecewise constant approximations with sizable intervals. By shedding light on these challenges, we advocate a refined understanding of Transformers' capabilities.
  arXiv  Detail & Related papers  (2024-02-23T18:12:53Z)
• Hiera: A Hierarchical Vision Transformer without the Bells-and-Whistles [65.54857068975068]
  In this paper, we argue that this additional bulk is unnecessary. By pretraining with a strong visual pretext task (MAE), we can strip out all the bells-and-whistles from a state-of-the-art multi-stage vision transformer. We create Hiera, an extremely simple hierarchical vision transformer that is more accurate than previous models.
  arXiv  Detail & Related papers  (2023-06-01T17:59:58Z)
• An Introduction to Transformers [23.915718146956355]
  transformer is a neural network component that can be used to learn useful sequences or sets of data-points. In this note we aim for a mathematically precise, intuitive, and clean description of the transformer architecture.
  arXiv  Detail & Related papers  (2023-04-20T14:54:19Z)
• What Makes for Good Tokenizers in Vision Transformer? [62.44987486771936]
  transformers are capable of extracting their pairwise relationships using self-attention. What makes for a good tokenizer has not been well understood in computer vision. Modulation across Tokens (MoTo) incorporates inter-token modeling capability through normalization. Regularization objective TokenProp is embraced in the standard training regime.
  arXiv  Detail & Related papers  (2022-12-21T15:51:43Z)
• Transformers learn in-context by gradient descent [58.24152335931036]
  Training Transformers on auto-regressive objectives is closely related to gradient-based meta-learning formulations. We show how trained Transformers become mesa-optimizers i.e. learn models by gradient descent in their forward pass.
  arXiv  Detail & Related papers  (2022-12-15T09:21:21Z)
• On the Power of Saturated Transformers: A View from Circuit Complexity [87.20342701232869]
  We show that saturated transformers transcend the limitations of hard-attention transformers. The jump from hard to saturated attention can be understood as increasing the transformer's effective circuit depth by a factor of $O(log n)$.
  arXiv  Detail & Related papers  (2021-06-30T17:09:47Z)
• A Survey on Visual Transformer [126.56860258176324]
  Transformer is a type of deep neural network mainly based on the self-attention mechanism. In this paper, we review these vision transformer models by categorizing them in different tasks and analyzing their advantages and disadvantages.
  arXiv  Detail & Related papers  (2020-12-23T09:37:54Z)

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.