Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
MOMENT: A Family of Open Time-series Foundation Models, ICML'24
| Date | Stars |
|---|---|
| 2026-07-31 | 812 |
| 2026-08-06 | 821 |
Today
+9 stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
<div align="center"> <img width="60%" alt="MOMENT" src="assets/MOMENT Logo.png"> <h1>MOMENT: A Family of Open Time-series Foundation Models</h1> [](https://arxiv.org/abs/2402.03885) [](https://huggingface.co/AutonLab/MOMENT-1-large) [](https://huggingface.co/datasets/AutonLab/Timeseries-PILE) [](https://opensource.org/license/MIT) []() </div> ## 🔥 News - Interested in LLM Agents for (Time Series) Machine Learning Engineering? Check out our latest work [TimeSeriesGym: A Scalable Benchmark for Time Series Machine Learning Engineering Agents](https://github.com/moment-timeseries-foundation-model/TimeSeriesGym) - We just released the [small](https://huggingface.co/AutonLab/MOMENT-1-small) and [base](https://huggingface.co/AutonLab/MOMENT-1-base) versions of the MOMENT model. - 🔥🔥🔥 We released [MOMENT research](https://github.com/moment-timeseries-foundation-model/moment-research) code, so you can pre-train your own time series foundation model, with your own data, and reproduce experiments from [our paper](https://arxiv.org/abs/2402.03885)! - We fixed an issue with Classification where MOMENT was unable to handle multi-channel inputs. - MOMENT was accepted at ICML 2024! - Interested in multimodal time series & text foundation models? Check out our preliminary work on JoLT (**Jo**intly **L**earned Represenations for **T**ime series & **T**ext) [[AAAI 2024 Student Abstract](https://ojs.aaai.org/index.php/AAAI/article/view/30423), [NeurIPS 2023 DGM4H Workshop](https://openreview.net/forum?id=UVF1AMBj9u)]. JoLT won the best student abstract presentation at AAAI! Stay tuned for multimodal time series & text foundation models! ## 📖 Introduction We introduce MOMENT, a family of open-source foundation models for general-purpose time-series analysis. Pre-training large models on time-series data is challenging due to (1) the absence a large and cohesive public time-series repository, and (2) diverse time-series characteristics which make multi-dataset training onerous. Additionally, (3) experimental benchmarks to evaluate these models especially in scenarios with limited resources, time, and supervision, are still in its nascent stages. To address these challenges, we compile a large and diverse collection of public time-series, called the Time-series Pile, and systematically tackle time-series-specific challenges to unlock large-scale multi-dataset pre-training. Finally, we build on recent work to design a benchmark to evaluate time-series foundation models on diverse tasks and datasets in limited supervision settings. Experiments on this benchmark demonstrate the effectiveness of our pre-trained models with minimal data and task-specific fine-tuning. Finally, we present several interesting empirical observations about large pre-trained time-series models. ### MOMENT: One Model, Multiple Tasks, Datasets & Domains <div align="center"> <img width="60%" alt="MOMENT: One Model, Multiple Tasks, Datasets & Domains" src="https://github.com/moment-timeseries-foundation-model/moment/assets/26150479/90c7d055-36d2-42aa-92b1-c5cfade22b3e"> </div> MOMENT on different datasets and tasks, without any parameter updates: - _Imputation:_ Better than statistical imputation baselines - _Anomaly Detection:_ Second best $F_1$ than all baselines - _Classification:_ More accurate than 11 / 16 compared methods - _Short-horizon Forecasting:_ Better than ARIMA on some datasets By linear probing (fine-tuning the final linear layer): - _Imputation:_ Better than baselines on 4 / 6 datasets - _Anomaly Detection:_ Best $F_1$ - _Long-horizon Forecasting:_ Competi
Excerpt of 14,731 characters
Read on GitHub
73
14
13
3
Ikko Eltociear Ashimine · Japan
1
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:c47bd32936d6cda7, topic:large-language-models