MONSTER

Monash Scalable Time Series Evaluation Repository

Angus Dempster*, Navid Mohammadi Foumani*, Chang Wei Tan,
Lynn Miller*, Amish Mishra*, Mahsa Salehi*,
Charlotte Pelletier†, Daniel Schmidt*, Geoffrey I. Webb*

*Monash University, Melbourne, Australia     †Université Bretagne Sud, IRISA, Vannes, France

Code on GitHub arXiv Paper Dataset on HuggingFace
MONSTER is a large-scale benchmark suite for time series classification. While UCR and UEA benchmarks offer common ground, they are limited by small dataset sizes—favoring models that reduce variance over those that scale.

Our goal is to expand the landscape with datasets that challenge models to be scalable, robust, and practical for real-world time series applications.

Please cite as:

@article{dempster_etal_2025,
  author  = {Dempster, Angus and Foumani, Navid Mohammadi and Tan, Chang Wei and Miller, Lynn and Mishra, Amish and Salehi, Mahsa and Pelletier, Charlotte and Schmidt, Daniel F and Webb, Geoffrey I},
  title   = {MONSTER: Monash Scalable Time Series Evaluation Repository},
  year    = {2025},
  journal = {arXiv:2502.15122},
}

Downloading Data

hf_hub_download

from huggingface_hub import hf_hub_download

path = hf_hub_download(repo_id = "monster-monash/Pedestrian", filename = "Pedestrian_X.npy", repo_type = "dataset")
X = np.load(path, mmap_mode = "r")

load_data

from datasets import load_dataset

dataset = load_dataset("monster-monash/Pedestrian", "fold_0", trust_remote_code = True)

Run

To view all command-line options:

python experiments/demo.py --help

Edit training parameters in experiments.py. For example:

self.parser.add_argument('--epochs', type=int, default=100, help='Number of training epochs')

To run on a specific dataset:

python experiments/demo.py --dataset UCIActivity

Project Folder Structure

. ├── experiments # Contains experimental scripts and demos │ └── demo.py # Example demo for running the models ├── figs # │ └── Logo.png # Project logo ├── models # Includes model definitions │ ├── deep_learning # Deep learning models for time series tasks │ │ ├── ConvTran.py # ConvTran model for time series classification │ │ └── FCN.py # Fully Convolutional Network (FCN) │ ├── loss.py # Loss functions │ ├── model_factory.py # Model factory for building models │ └── non_deep # Non-deep learning models │ ├── hydra_gpu.py # │ ├── quant.py # │ ├── ridge.py # │ └── utils.py # ├── notebook # Jupyter notebooks ├── README.md # ├── requirements.txt # ├── results # Outputs and logs ├── src # Source code │ ├── analysis.py # │ ├── data_loader.py # │ ├── trainer.py # │ └── utils.py # └── .gitignore #