Forecasting as Rendering: A 2D Gaussian Splatting Framework for Time Series Forecasting

2026-02-10

AI Summary

This work addresses the challenges in time series forecasting arising from the complex coupling between intra-period fluctuations and inter-period trends, the disruption of temporal continuity caused by conventional 2D representations, and the inflexibility of fixed resolution in capturing non-stationary patterns. To overcome these issues, the authors reformulate the forecasting task as a 2D generative rendering problem, modeling the future sequence as an implicit two-dimensional temporal surface. The key innovations include a Multi-Basis Gaussian Kernel Generation (MB-GKG) module that enables anisotropic and adaptive geometric alignment, and a Multi-Period Temporally Continuous Rasterization (MP-CCR) mechanism that preserves temporal continuity across period boundaries. Extensive experiments on multiple standard benchmarks demonstrate that the proposed method achieves state-of-the-art or highly competitive performance, confirming its effectiveness and generalization capability.

Abstract

Time series forecasting (TSF) remains a challenging problem due to the intricate entanglement of intraperiod-fluctuations and interperiod-trends. While recent advances have attempted to reshape 1D sequences into 2D period-phase representations, they suffer from two principal limitations.Firstly, treating reshaped tensors as static images results in a topological mismatch, as standard spatial operators sever chronological continuity at grid boundaries. Secondly, relying on uniform fixed-size representations allocates modeling capacity inefficiently and fails to provide the adaptive resolution required for compressible, non-stationary temporal patterns. To address these limitations, we introduce TimeGS, a novel framework that fundamentally shifts the forecasting paradigm from regression to 2D generative rendering. By reconceptualizing the future sequence as a continuous latent surface, TimeGS utilizes the inherent anisotropy of Gaussian kernels to adaptively model complex variations with flexible geometric alignment. To realize this, we introduce a Multi-Basis Gaussian Kernel Generation (MB-GKG) block that synthesizes kernels from a fixed dictionary to stabilize optimization, and a Multi-Period Chronologically Continuous Rasterization (MP-CCR) block that enforces strict temporal continuity across periodic boundaries. Comprehensive experiments on standard benchmark datasets demonstrate that TimeGS attains state-of-the-art performance.

Problem

• time series forecasting
• 2D representation
• temporal continuity
• non-stationary patterns
• topological mismatch

Innovation

• 2D Gaussian Splatting
• Time Series Forecasting
• Generative Rendering
• Chronological Continuity
• Adaptive Resolution