SurF: A Generative Model for Multivariate Irregular Time Series Forecasting
cs.LG
Submitted: 2026-05-13
Updated: 2026-09-09
Code: https://github.com/MrRezaeiUofT/SurF
Project page: https://nijianmo.github.io/amazon
License: http://creativecommons.org/licenses/by/4.0/
The gist: Irregularly sampled multivariate event streams remain a difficult modality for generative modeling: tokenization-based approaches break down when inter-event intervals vary by orders of magnitude.
Terminology
Abstract
Irregularly sampled multivariate event streams remain a difficult modality for generative modeling: tokenization-based approaches break down when inter-event intervals vary by orders of magnitude. We (i) propose SurF, a generative model that uses the Time Rescaling Theorem (TRT) as a learnable bijection between event sequences and i.i.d. unit-rate exponential noise, enabling a single model to be trained across heterogeneous event-stream datasets; (ii) three efficient parameterizations of the cumulative intensity that scale to long sequences; and (iii) a Transformer-based encoder for multi-dataset pretraining. On six real-world benchmarks, SurF achieves the best reported time RMSE on Earthquake, Retweet, and Taobao, and is within trial-level noise of the strongest specialist on the remaining three. Under a strict leave-one-out protocol, the held-out checkpoint beats every classical and neural-autoregressive baseline on 5/6 datasets and beats every baseline on Amazon and Earthquake, an initial step toward foundation models over asynchronous event streams (Code is available at https://github.com/MrRezaeiUofT/SurF).
Sources
- A decoder-only foundation model for time-series forecasting
- Chronos: Learning the Language of Time Series
- Lag-Llama: Towards Foundation Models for Probabilistic Time Series Forecasting
- Multi-Time Attention Networks for Irregularly Sampled Time Series
- Fully Neural Network based Model for General Temporal Point Processes
- Intensity-Free Learning of Temporal Point Processes
- The Neural Hawkes Process: A Neurally Self-Modulating Multivariate Point Process
- Transformer Hawkes Process
- HYPRO: A Hybridly Normalized Probabilistic Model for Long-Horizon Prediction of Event Sequences
- Transformer Embeddings of Irregularly Spaced Events and Their Participants
- Still Competitive: Revisiting Recurrent Models for Irregular Time Series Prediction
- LLaMA: Open and Efficient Foundation Language Models
- DINOv2: Learning Robust Visual Features without Supervision
- Neural Temporal Point Processes: A Review
- EasyTPP: Towards Open Benchmarking Temporal Point Processes
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks