Distance Metric Learning Algorithms for Python

Many machine learning algorithms need a similarity measure to carry out their tasks. Usually, standard distances, like euclidean distance, are used to measure this similarity. Distance Metric Learning algorithms try to learn an optimal distance from the data.

There are two main ways to learn a distance in Distance Metric Learning:

• Learning a metric matrix M, that is, a positive semidefinite matrix. In this case, the distance is measured as

• Learning a linear map L. This map is also represented by a matrix, not necessarily definite or squared. Here, the distance between two elements is the euclidean distance after applying the transformation.

Every linear map defines a single metric (M = L'L), and two linear maps that define the same metric only differ in an isometry. So both approaches are equivalent.

Improving 1-NN classification.

Learning a projection onto a plane for the digits dataset (dimension 64).

See the available algorithms, the additional functionalities and the full documentation here.

If you find this package useful in your research, please consider citing the software:

@article{suarez2020pydml,
  title={pyDML: A Python Library for Distance Metric Learning},
  author={Su{\'a}rez, Juan Luis and Garc{\'i}a, Salvador and Herrera, Francisco},
  journal={Journal of Machine Learning Research},
  volume={21},
  number={96},
  pages={1--7},
  year={2020}
}

The distance metric learning algorithms in pyDML are being evaluated in several datasets. The results of these experiments are available in the pyDML-Stats repository.

• PyPI latest version: pip install pyDML

• From GitHub: clone or download this repository and run the command python setup.py install on the root directory.

• Juan Luis Suárez Díaz (jlsuarezdiaz)