Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Position: The Inevitable End of One-Architecture-Fits-All-Domains in Time Series Forecasting

Qinwei Ma, Jingzhe Shi, Jiahao Qiu, Zaiwen Yang

TL;DR

This paper argues that time series forecasting cannot be effectively solved with a single universal architecture due to inherent cross-domain heterogeneity and the temporal limits of data. It analyzes the root causes—cross-domain diversity and data scarcity—and documents how general-domain TSF progress often fails to translate to domain-specific ceilings, with evidence from simple baselines beating complex models and saturation in benchmarks. The authors propose two primary directions: develop domain-specific neural networks (DSNNs) that encode domain priors, or adopt meta-learning systems (including agentic LLM-based orchestration and meta-selection) to dynamically deploy the right expert for a given domain. They also discuss alternative views, acknowledging the value of general SOTA benchmarks while arguing for a practical shift toward domain-aware strategies to improve performance in high-stakes, real-world forecasting settings.

Abstract

Recent work has questioned the effectiveness and robustness of neural network architectures for time series forecasting tasks. We summarize these concerns and analyze groundly their inherent limitations: i.e. the irreconcilable conflict between single (or few similar) domains SOTA and generalizability over general domains for time series forecasting neural network architecture designs. Moreover, neural networks architectures for general domain time series forecasting are becoming more and more complicated and their performance has almost saturated in recent years. As a result, network architectures developed aiming at fitting general time series domains are almost not inspiring for real world practices for certain single (or few similar) domains such as Finance, Weather, Traffic, etc: each specific domain develops their own methods that rarely utilize advances in neural network architectures of time series community in recent 2-3 years. As a result, we call for the time series community to shift focus away from research on time series neural network architectures for general domains: these researches have become saturated and away from domain-specific SOTAs over time. We should either (1) focus on deep learning methods for certain specific domain(s), or (2) turn to the development of meta-learning methods for general domains.

Position: The Inevitable End of One-Architecture-Fits-All-Domains in Time Series Forecasting

TL;DR

This paper argues that time series forecasting cannot be effectively solved with a single universal architecture due to inherent cross-domain heterogeneity and the temporal limits of data. It analyzes the root causes—cross-domain diversity and data scarcity—and documents how general-domain TSF progress often fails to translate to domain-specific ceilings, with evidence from simple baselines beating complex models and saturation in benchmarks. The authors propose two primary directions: develop domain-specific neural networks (DSNNs) that encode domain priors, or adopt meta-learning systems (including agentic LLM-based orchestration and meta-selection) to dynamically deploy the right expert for a given domain. They also discuss alternative views, acknowledging the value of general SOTA benchmarks while arguing for a practical shift toward domain-aware strategies to improve performance in high-stakes, real-world forecasting settings.

Abstract

Recent work has questioned the effectiveness and robustness of neural network architectures for time series forecasting tasks. We summarize these concerns and analyze groundly their inherent limitations: i.e. the irreconcilable conflict between single (or few similar) domains SOTA and generalizability over general domains for time series forecasting neural network architecture designs. Moreover, neural networks architectures for general domain time series forecasting are becoming more and more complicated and their performance has almost saturated in recent years. As a result, network architectures developed aiming at fitting general time series domains are almost not inspiring for real world practices for certain single (or few similar) domains such as Finance, Weather, Traffic, etc: each specific domain develops their own methods that rarely utilize advances in neural network architectures of time series community in recent 2-3 years. As a result, we call for the time series community to shift focus away from research on time series neural network architectures for general domains: these researches have become saturated and away from domain-specific SOTAs over time. We should either (1) focus on deep learning methods for certain specific domain(s), or (2) turn to the development of meta-learning methods for general domains.
Paper Structure (26 sections, 1 theorem, 2 equations, 3 figures, 2 tables)

This paper contains 26 sections, 1 theorem, 2 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Neural Network Methods for General-domain Time Series Forecasting
  4. Large Multi-Domain Time Series Datasets and Large Foundational Time Series Models
  5. The Irreconcilable Conflict: Single-Domain vs. General-Domain SOTA for Time Series Neural Networks
  6. Cross-Domain Diversity
  7. Data Scarcity
  8. Natural Scarcity and Augmentation Challenge
  9. The Temporal Boundary of Approximation Error
  10. Failure of Scaling and the Need of Human Priors
  11. Consequences of the conflict and focusing on a 'General Domain Time Series Forecasting'
  12. Peculiar observations in Time Series Forecasting Community
  13. Diverse SOTA proposed by different task Communities irrelevant to the time series community
  14. Possible Solutions
  15. Domain-specific Neural Network
  16. ...and 11 more sections

Key Result

Theorem 1.1

Let $L$ be a loss function bounded by $M$ and $H$ be any set of hypotheses functions. Suppose $T=ma$ for some $m,a > 0$. Then for any $\delta > a(m - 1)\beta(a)$ with probability $1 - \delta$: holds for arbitrary hypothesis function $h \in H$, where $\delta^{\prime}=\delta-a(m-1)\mathsf{\beta}(a)$ and $\mathfrak{R}_m(H,\Pi)=\frac{1}{m}\mathbb{E}[\sup_{h\in H}\sum_{i=1}^m\sigma_i\ell(h,\widetilde{

Figures (3)

  • Figure 1: Left: A mind map of the core arguments in this position paper. We propose that there exists a inherent, irreconcilable conflict between general-domain and single-domain performance for time series forecasting neural network architecture designing. We analyze its causes, consequences and advocated better research directions. We discuss more about Alternative Views in Section \ref{['sec: alternative views']}. Right: Error reached/reachable for a specific domain (e.g. weather forecasting) when utilizing different types of methods with different available domain-specific context and properties.
  • Figure 2: Left: Number of total timestamps across different domains for multivariate forecasting datasets in TFB dataset tfbbenchmark; Right: Number of total timestamps across different domains forecasting tasks used by most TSF papers represented by ModernTCN moderntcn. All are small compared to typical NLP datasets gpt.
  • Figure 3: Suggested practices: from pursuing a neural network architecture that has huge gaps with domain SOTAs to (1) developing single-domain SOTA neural network methods, or (2) developing meta-learning methods that are generalizable across domains.

Theorems & Definitions (2)

  • Theorem 1.1
  • proof