Cover Image

Curse of Attention: A Kernel-Based Perspective for Why Transformers Fail to Generalize on Time Series Forecasting and Beyond

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ke, Yekun, Liang, Yingyu, Shi, Zhenmei, Song, Zhao, Yang, Chiwun
Format:	Preprint
Published:	2024
Subjects:	Machine Learning Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Towards Infinite-Long Prefix in Transformer
by: Liang, Yingyu, et al.
Published: (2024)

Provable Differentially Private Computation of the Cross-Attention Mechanism
by: Ke, Yekun, et al.
Published: (2024)

Visual Autoregressive Transformers Must Use $Ω(n^2 d)$ Memory
by: Cao, Yang, et al.
Published: (2025)

Advancing the Understanding of Fixed Point Iterations in Deep Neural Networks: A Detailed Analytical Study
by: Ke, Yekun, et al.
Published: (2024)

Circuit Complexity Bounds for Visual Autoregressive Model
by: Ke, Yekun, et al.
Published: (2025)

Training Tensor Attention Efficiently: From Cubic to Almost Linear Time
by: Cao, Yang, et al.
Published: (2024)

Exploring the Frontiers of Softmax: Provable Optimization, Applications in Diffusion Model, and Beyond
by: Cao, Yang, et al.
Published: (2024)

On Computational Limits and Provably Efficient Criteria of Visual Autoregressive Models: A Fine-Grained Complexity Analysis
by: Ke, Yekun, et al.
Published: (2025)

Theoretical Foundation of Flow-Based Time Series Generation: Provable Approximation, Generalization, and Efficiency
by: Long, Jiangxuan, et al.
Published: (2025)

Beyond Linear Approximations: A Novel Pruning Approach for Attention Matrix
by: Liang, Yingyu, et al.
Published: (2024)

RoPE Attention Can Be Trained in Almost Linear Time
by: Cao, Yang, et al.
Published: (2024)

Differential Privacy Mechanisms in Neural Tangent Kernel Regression
by: Gu, Jiuxiang, et al.
Published: (2024)

Multi-Layer Transformers Gradient Can be Approximated in Almost Linear Time
by: Liang, Yingyu, et al.
Published: (2024)

Theoretical Constraints on the Expressive Power of $\mathsf{RoPE}$-based Tensor Attention Transformers
by: Li, Xiaoyu, et al.
Published: (2024)

Conv-Basis: A New Paradigm for Efficient Attention Inference and Gradient Computation in Transformers
by: Liang, Yingyu, et al.
Published: (2024)

Unraveling the Smoothness Properties of Diffusion Models: A Gaussian Mixture Perspective
by: Liang, Yingyu, et al.
Published: (2024)

When Can We Solve the Weighted Low Rank Approximation Problem in Truly Subquadratic Time?
by: Li, Chenyang, et al.
Published: (2025)

HSR-Enhanced Sparse Attention Acceleration
by: Chen, Bo, et al.
Published: (2024)

Why Do Transformers Fail to Forecast Time Series In-Context?
by: Zhou, Yufa, et al.
Published: (2025)

Scaling Law Phenomena Across Regression Paradigms: Multiple and Kernel Approaches
by: Chen, Yifang, et al.
Published: (2025)

On Computational Limits of FlowAR Models: Expressivity and Efficiency
by: Cao, Yang, et al.
Published: (2025)

Why Larger Language Models Do In-context Learning Differently?
by: Shi, Zhenmei, et al.
Published: (2024)

Bypassing the Exponential Dependency: Looped Transformers Efficiently Learn In-context by Multi-step Gradient Descent
by: Chen, Bo, et al.
Published: (2024)

Unifying Learning Dynamics and Generalization in Transformers Scaling Law
by: Yang, Chiwun
Published: (2025)

Neural Algorithmic Reasoning for Hypergraphs with Looped Transformers
by: Huang, Zekai, et al.
Published: (2025)

HOFAR: High-Order Augmentation of Flow Autoregressive Transformers
by: Liang, Yingyu, et al.
Published: (2025)

Time and Memory Trade-off of KV-Cache Compression in Tensor Transformer Decoding
by: Chen, Yifang, et al.
Published: (2025)

Kernel Regression in Structured Non-IID Settings: Theory and Implications for Denoising Score Learning
by: Zhang, Dechen, et al.
Published: (2025)

Universal Approximation of Visual Autoregressive Transformers
by: Chen, Yifang, et al.
Published: (2025)

On Fine-Grained I/O Complexity of Attention Backward Passes
by: Li, Xiaoyu, et al.
Published: (2024)

A Tighter Complexity Analysis of SparseGPT
by: Li, Xiaoyu, et al.
Published: (2024)

On the Computational Capability of Graph Neural Networks: A Circuit Complexity Bound Perspective
by: Li, Xiaoyu, et al.
Published: (2025)

Do Large Language Models Have Compositional Ability? An Investigation into Limitations and Scalability
by: Xu, Zhuoyan, et al.
Published: (2024)

Circuit Complexity Bounds for RoPE-based Transformer Architecture
by: Chen, Bo, et al.
Published: (2024)

How Sparse Attention Approximates Exact Attention? Your Attention is Naturally $n^C$-Sparse
by: Deng, Yichuan, et al.
Published: (2024)

Looped ReLU MLPs May Be All You Need as Practical Programmable Computers
by: Liang, Yingyu, et al.
Published: (2024)

The Computational Limits of State-Space Models and Mamba via the Lens of Circuit Complexity
by: Chen, Yifang, et al.
Published: (2024)

Modern Hopfield Networks Require Chain-of-Thought to Solve $\mathsf{NC}^1$-Hard Problems
by: Cao, Yang, et al.
Published: (2024)

Decentralized Attention Fails Centralized Signals: Rethinking Transformers for Medical Time Series
by: Yu, Guoqi, et al.
Published: (2026)

Attention as Robust Representation for Time Series Forecasting
by: Niu, PeiSong, et al.
Published: (2024)