pyDML: A Python Library for Distance Metric Learning
Authors: Juan Luis Suárez, Salvador García, Francisco Herrera
JMLR 2020 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The py DML package currently provides more than 20 algorithms... In addition, the library also provides some utilities for the visualization of classifier regions, parameter tuning and a stats website with the performance of the implemented algorithms. The py DML library also incorporates graphical tools for the representation and evaluation of the learned distances... Finally, a stats website is also provided, where the performance of the implemented algorithms is evaluated under different conditions (Suarez et al., 2019). |
| Researcher Affiliation | Academia | Juan Luis Su arez EMAIL Salvador Garc ıa EMAIL Francisco Herrera EMAIL Da SCI, Andalusian Research Institute in Data Science and Computational Intelligence University of Granada, Granada, Spain |
| Pseudocode | No | The paper includes a code snippet in Figure 1, but it is an example of how to use the library's functions in Python, not a pseudocode or algorithm block detailing the logic of any of the DML algorithms themselves. |
| Open Source Code | Yes | Source code and documentation can be found at https://github.com/jlsuarezdiaz/py DML. |
| Open Datasets | Yes | Figure 1 shows a basic example... >>> from sklearn.datasets import load_iris # Iris dataset |
| Dataset Splits | No | The paper mentions 'tune functions, which allow the parameters of the DML algorithms to be easily estimated with cross validation', implying the use of cross-validation for parameter tuning. However, it does not specify any particular dataset splits (e.g., percentages, folds) for general experimental reproduction within the paper's main text. |
| Hardware Specification | No | The paper describes a software library and its features, compatibility, and installation. It does not provide any specific hardware details (like CPU, GPU models, or memory) used for developing, testing, or evaluating the library or its algorithms within the paper's content. |
| Software Dependencies | No | The package relies on the scipy ecosystem, it is fully compatible with scikit-learn, and is distributed under GPLv3 license... The main one is Scikit-Learn (Pedregosa et al., 2011), an efficient open-source library for machine learning, which relies on the Scipy ecosystem, which contains numerical calculus libraries, such as Num Py, data processing libraries, such as Pandas, or data visualization libraries, such as Matplotlib. These software components are mentioned, but specific version numbers are not provided for them. |
| Experiment Setup | No | The paper states, 'It is important to emphasize that these algorithms include different hyperparameters that can be modified to improve the performance or to change the conditions of the learned distances. To this end the package includes tune functions, which allow the parameters of the DML algorithms to be easily estimated with cross validation...' While it mentions hyperparameters and tuning, it does not provide specific values for any experimental setup (e.g., learning rates, batch sizes, epochs) for any experiment detailed within the paper itself. |