aeon: a Python Toolkit for Learning from Time Series
Authors: Matthew Middlehurst, Ali Ismail-Fawaz, Antoine Guillaume, Christopher Holder, David Guijo-Rubio, Guzal Bulatova, Leonidas Tsaprounis, Lukasz Mentel, Martin Walter, Patrick Schäfer, Anthony Bagnall
JMLR 2024 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | The paper describes 'aeon: a Python Toolkit for Learning from Time Series' and its design, implementation, and features. While the toolkit is designed for experimental research in machine learning, the paper itself does not present empirical studies, data analysis, performance metrics, or hypothesis validation. It focuses on describing the software artifact rather than conducting experiments. |
| Researcher Affiliation | Collaboration | 1ECS, University of Southampton, United Kingdom 2IRIMAS, Universit e de Haute-Alsace, France 3Novah e and Constellation, France 4CMP, University of East Anglia, United Kingdom 5DIAN, University of C ordoba, Spain 6Humboldt-Universit at zu Berlin, Germany The affiliations include universities (academic) such as the University of Southampton and Humboldt-Universität zu Berlin, and a company (industry) 'Novahé and Constellation'. This mix indicates a collaborative affiliation. |
| Pseudocode | No | The paper includes Python code examples to demonstrate the usage of the aeon toolkit, such as loading datasets and fitting models. However, it does not contain any structured pseudocode or algorithm blocks (e.g., labeled 'Algorithm 1' or similar code-like procedures). |
| Open Source Code | Yes | The package is distributed under the 3-Clause BSD license and is available at https://github.com/ aeon-toolkit/aeon. |
| Open Datasets | Yes | from aeon.datasets import load_airline from aeon.datasets import load_classification The paper demonstrates the use of datasets through the `aeon.datasets` module, which provides access to datasets for time series tasks, indicating that these datasets are accessible via the toolkit. |
| Dataset Splits | Yes | train_X , train_y , _ = load_classification ("Gun Point", split="Train") test_X , _, _ = load_classification ("Gun Point", split="Test") This code snippet explicitly shows the use of 'split="Train"' and 'split="Test"' parameters for loading data, indicating predefined dataset splits. |
| Hardware Specification | No | No specific hardware details such as CPU/GPU models, memory specifications, or cloud computing instances used for running experiments are mentioned in the paper. |
| Software Dependencies | No | aeon supports all versions from Python 3.8 onwards in version v0.5.0. The primary dependency is scikit-learn, and the package contains extensive usage of the numpy and scipy libraries. aeon depends on numba and also includes optional/soft dependencies to packages such as statsmodels, tensorflow, and tsfresh. While Python 3.8 is mentioned, no specific version numbers are provided for the other key software dependencies like scikit-learn, numpy, scipy, numba, statsmodels, tensorflow, or tsfresh. A programming language version alone is insufficient without versioned libraries. |
| Experiment Setup | No | The paper provides code examples demonstrating basic usage of the toolkit, such as `forecaster = TrendForecaster()` and `forecaster.fit(y)`. However, it does not provide specific experimental setup details like hyperparameter values (e.g., learning rates, batch sizes, epochs), optimizer settings, or training schedules. |