Joint Dimensionality Reduction and Metric Learning: A Geometric Take

Abstract

To be tractable and robust to data noise, existing metric learning algorithms commonly rely on PCA as a pre-processing step. How can we know, however, that PCA, or any other specific dimensionality reduction technique, is the method of choice for the problem at hand? The answer is simple: We cannot! To address this issue, in this paper, we develop a Riemannian framework to jointly learn a mapping performing dimensionality reduction and a metric in the induced space. Our experiments evidence that, while we directly work on high-dimensional features, our approach yields competitive runtimes with and higher accuracy than state-of-the-art metric learning algorithms.

Cite

Text

Harandi et al. "Joint Dimensionality Reduction and Metric Learning: A Geometric Take." International Conference on Machine Learning, 2017.

Markdown

[Harandi et al. "Joint Dimensionality Reduction and Metric Learning: A Geometric Take." International Conference on Machine Learning, 2017.](https://mlanthology.org/icml/2017/harandi2017icml-joint/)

BibTeX

@inproceedings{harandi2017icml-joint,
  title     = {{Joint Dimensionality Reduction and Metric Learning: A Geometric Take}},
  author    = {Harandi, Mehrtash and Salzmann, Mathieu and Hartley, Richard},
  booktitle = {International Conference on Machine Learning},
  year      = {2017},
  pages     = {1404-1413},
  volume    = {70},
  url       = {https://mlanthology.org/icml/2017/harandi2017icml-joint/}
}