https://github.com/SaberaTalukder/TOTEM

The official code 👩‍💻 for - TOTEM: TOkenized Time Series EMbeddings for General Time Series Analysis
https://github.com/SaberaTalukder/TOTEM

foundation-models representation-learning time-series time-series-analysis time-series-anomaly-detection time-series-forecasting time-series-foundation-model time-series-imputation tokenization

Last synced: 23 days ago
JSON representation

The official code 👩‍💻 for - TOTEM: TOkenized Time Series EMbeddings for General Time Series Analysis

• Host: GitHub
• URL: https://github.com/SaberaTalukder/TOTEM
• Owner: SaberaTalukder
• Created: 2024-02-27T09:06:34.000Z (7 months ago)
• Default Branch: master
• Last Pushed: 2024-02-27T09:35:06.000Z (7 months ago)
• Last Synced: 2024-05-15T20:54:02.630Z (4 months ago)
• Topics: foundation-models, representation-learning, time-series, time-series-analysis, time-series-anomaly-detection, time-series-forecasting, time-series-foundation-model, time-series-imputation, tokenization
• Language: Python
• Homepage: https://arxiv.org/pdf/2402.16412.pdf
• Size: 58.6 KB
• Stars: 46
• Watchers: 4
• Forks: 4
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# TOTEM: TOkenized Time series EMbeddings for General Time Series Analysis

TOTEM explores time series unification through discrete tokens (not patches!!). Its simple VQVAE backbone learns a self-supervised, discrete, codebook in either a generalist (multiple domains) or specialist (1 domain) manner.

TOTEM's codebook can then be tested on in domain or zero shot data with many 🔥 time series tasks.

Check out the [paper](https://arxiv.org/pdf/2402.16412.pdf) for more details!

## Get Started with TOTEM 💪

### 1. Setup your environment 🤓

```

pip install -r requirements.txt

```

### 2. Get the [data](https://drive.google.com/drive/u/0/folders/1gI36rS8irRZ32ibzKBPGncDmMXQtEf1C) ⏳

### 3. Run TOTEM 🚀

```

# Imputation Specialist

imputation/scripts/electricity.sh or ETTh1.sh or ETTh2.sh or ETTm1.sh or ETTm2.sh or weather.sh

# Imputation Generalist

imputation/scripts/all.sh

# Anomaly Detection Specialist

anomaly_detection/scripts/msl.sh or psm.sh or smap.sh or smd.sh or swat.sh

# Anomaly Detection Generalist

anomaly_detection/scripts/all.sh

# Forecasting Specialist

forecasting/scripts/electricity.sh or ETTh1.sh or ETTh2.sh or ETTm1.sh or ETTm2.sh or weather.sh or traffic.sh

# Forecasting Generalist

forecasting/scripts/all.sh

# Process Zero Shot Data

process_zero_shot_data/scripts/neuro2.sh or neuro5.sh or saugeen.sh or sunspot.sh or us_births.sh

```

### 4. Model Zoo (a.k.a Pretrained Models) 🦑🐯🐊🐳

Coming Soon!

## Cite If You ❤️ TOTEM

```

@misc{talukder2024totem,

      title={TOTEM: TOkenized Time Series EMbeddings for General Time Series Analysis}, 

      author={Sabera Talukder and Yisong Yue and Georgia Gkioxari},

      year={2024},

      eprint={2402.16412},

      archivePrefix={arXiv},

      primaryClass={cs.LG}

}