Perplexity-free Parametric t-SNE
(2020) European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning (2.October.2020)
Files
Crecchi2020_parametric_MStSNE.pdf
Open Access - Adobe PDF
- 2.05 MB
Details
- Authors
De Bodt, Cyril 
- Abstract
- The t-distributed Stochastic Neighbor Embedding (t-SNE) algorithm is a ubiquitously employed dimensionality reduction (DR) method. Its non-parametric nature and impressive efficacy motivated its parametric extension. It is however bounded to a user-defined perplexity parameter, restricting its DR quality compared to recently developed multi-scale perplexity-free approaches. This paper hence proposes a multi-scale parametric t-SNE scheme, relieved from the perplexity tuning and with a deep neural network implementing the mapping. It produces reliable embeddings with out-of-sample extensions, competitive with the best perplexity adjustments in terms of neighborhood preservation on multiple data sets.
- Affiliations
Citations
Crecchi Francesco, De Bodt, C., Verleysen, M., Lee, J., & Bacciu Davide. (2020). Perplexity-free Parametric t-SNE. ESANN 2020 proceedings, p. 387-392. https://hdl.handle.net/2078.5/219779