English
Related papers

Related papers: Transformer Approximations from ReLUs

200 papers

We prove that with linear transformations, both (i) two-layer self-attention and (ii) one-layer self-attention followed by a softmax function are universal approximators for continuous sequence-to-sequence functions on compact domains. Our…

Machine Learning · Computer Science 2025-12-17 Jerry Yao-Chieh Hu , Hude Liu , Hong-Yu Chen , Weimin Wu , Han Liu

Previous research observed accuracy degradation when replacing the attention softmax with a point-wise activation such as ReLU. In the context of vision transformers, we find that this degradation is mitigated when dividing by sequence…

Computer Vision and Pattern Recognition · Computer Science 2023-10-18 Mitchell Wortsman , Jaehoon Lee , Justin Gilmer , Simon Kornblith

To enhance the computational efficiency of quantized Transformers, we replace the dot-product and Softmax-based attention with an alternative mechanism involving addition and ReLU activation only. This side-steps the expansion to double…

Machine Learning · Computer Science 2025-10-02 Rickard Brännvall , Andrei Stoian

The Transformer architecture consists of self-attention and feed-forward networks (FFNs) which can be viewed as key-value memories according to previous works. However, FFN and traditional memory utilize different activation functions…

Computation and Language · Computer Science 2023-02-14 Kai Shen , Junliang Guo , Xu Tan , Siliang Tang , Rui Wang , Jiang Bian

Formal verification of transformers has become increasingly important due to their widespread deployment in safety-critical applications. Compared to classic neural networks, the inferences of transformers involve highly complex…

Artificial Intelligence · Computer Science 2026-05-15 Hengjie Liu , Zhenya Zhang , Jianjun Zhao

Attention is a key part of the transformer architecture. It is a sequence-to-sequence mapping that transforms each sequence element into a weighted sum of values. The weights are typically obtained as the softmax of dot products between…

Transformers excel through content-addressable retrieval and the ability to exploit contexts of, in principle, unbounded length. We recast associative memory at the level of probability measures, treating a context as a distribution over…

Machine Learning · Statistics 2026-02-03 Ryotaro Kawata , Taiji Suzuki

Recently, it has been argued that encoder-decoder models can be made more interpretable by replacing the softmax function in the attention with its sparse variants. In this work, we introduce a novel, simple method for achieving sparsity in…

Computation and Language · Computer Science 2021-10-07 Biao Zhang , Ivan Titov , Rico Sennrich

Non-linear operations such as GELU, Layer normalization, and Softmax are essential yet costly building blocks of Transformer models. Several prior works simplified these operations with look-up tables or integer computations, but such…

Machine Learning · Computer Science 2021-12-07 Joonsang Yu , Junki Park , Seongmin Park , Minsoo Kim , Sihwa Lee , Dong Hyun Lee , Jungwook Choi

Transformers have improved drastically the performance of natural language processing (NLP) and computer vision applications. The computation of transformers involves matrix multiplications and non-linear activation functions such as…

Hardware Architecture · Computer Science 2024-02-19 Christodoulos Peltekis , Kosmas Alexandridis , Giorgos Dimitrakopoulos

Following up on the linear transformer part of the article from Katharopoulos et al., that takes this idea from Shen et al., the trick that produces a linear complexity for the attention mechanism is re-used and extended to a second-order…

Machine Learning · Computer Science 2020-10-29 Jean Mercat

We introduce sliced ReLU attention, a new attention mechanism that departs structurally from both softmax and its approximation alternatives. Instead of applying a nonlinearity to pairwise dot products, we operate on one-dimensional…

Machine Learning · Computer Science 2026-02-05 François-Xavier Vialard , Siwan Boufadène

The original softmax-based attention mechanism (regular attention) in the extremely successful Transformer architecture computes attention between $N$ tokens, each embedded in a $D$-dimensional head, with a time complexity of $O(N^2D)$.…

Machine Learning · Computer Science 2025-10-28 Armin Gerami , Ramani Duraiswami

This paper investigates the learning theory of Transformer networks for regression tasks on the compact Euclidean domain $[0,1]^d$ and $d$-dimensional compact Riemannian manifolds. We propose a novel constructive approximation framework for…

Machine Learning · Statistics 2026-05-12 Zhongjie Shi , Wenjing Liao

Large language models (LLMs), such as ChatGPT and GPT4, have shown outstanding performance in many human life task. Attention computation plays an important role in training LLMs. Softmax unit and ReLU unit are the key structure in…

Machine Learning · Computer Science 2023-08-17 Yichuan Deng , Zhao Song , Shenghao Xie

Various linear complexity models, such as Linear Transformer (LinFormer), State Space Model (SSM), and Linear RNN (LinRNN), have been proposed to replace the conventional softmax attention in Transformer structures. However, the optimal…

Machine Learning · Computer Science 2024-11-19 Yuhong Chou , Man Yao , Kexin Wang , Yuqi Pan , Ruijie Zhu , Yiran Zhong , Yu Qiao , Jibin Wu , Bo Xu , Guoqi Li

In this paper, we have extended the well-established universal approximator theory to neural networks that use the unbounded ReLU activation function and a nonlinear softmax output layer. We have proved that a sufficiently large neural…

Machine Learning · Computer Science 2020-02-12 Behnam Asadi , Hui Jiang

Large transformer models have achieved state-of-the-art results in numerous natural language processing tasks. Among the pivotal components of the transformer architecture, the attention mechanism plays a crucial role in capturing token…

Computation and Language · Computer Science 2026-03-16 Yichuan Deng , Zhao Song , Kaijun Yuan , Tianyi Zhou

The attention mechanism is the computational core of modern Transformer architectures, but its quadratic complexity in the input sequence length is the bottleneck for large-scale inference. This has motivated a rapidly growing body of work…

We explore the expressive power of Transformers by establishing precise approximation error upper and lower bounds for H\"{o}lder class. Specifically, a new approximation upper bound is derived for the standard Transformer architecture…

Machine Learning · Computer Science 2026-05-11 Xin He , Yuling Jiao , Xiliang Lu , Jerry Zhijian Yang
‹ Prev 1 2 3 10 Next ›