Quality assessment of dimensionality reduction: Rank-based criteria
By John Aldo Lee, Michel Verleysen. 2009
334 citations recorded by the source corpus.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Media Technology
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Sergey Levine
- Daniel Kappler
- John Aldo Lee
- Jeannette Bohg
- Michel Verleysen
- Cyril de Bodt
- Dounia Mulders
- Stefan Schaal
In Computer Vision and Pattern Recognition
- John Aldo Lee
- Michel Verleysen
- Cyril de Bodt
- Dounia Mulders
- Jeannette Bohg
- Daniel Kappler
- Stefan Schaal
- Felix Widmaier
In Media Technology
- Michel Verleysen
- John Aldo Lee
- Stefan Schaal
- Jeannette Bohg
- Daniel Kappler
- Felix Widmaier
- Yuanyuan Ren
- Peng Zhang
Citations in the corpus
- Toward a Quantitative Survey of Dimension Reduction Techniques (2019, methodology): “Projection quality is defined as a single scalar value using the rank-based criteria in [22].”
- Shift-invariant similarities circumvent distance concentration in stochastic neighbor embedding and variants (2011, methodology): “It varies between 0 and 1; for a random embedding, QNX(K) ≈ K/(N − 1) [31].”
- Robot arm pose estimation by pixel-wise regression of joint angles (2016, methodology): “10 shows an evaluation of the quality of the mappings by tSNE and KPCA using the QNX(K) measure as proposed in [32].”
- Perplexity-free t-SNE and twice Student tt-SNE (2018, background): “Evaluated in O N 2 log N time [18], its area 1] grows with DR quality, quantified at all scales with an emphasis on small ones.”
- Making machine learning models interpretable (2012, background): “Many of these come under the umbrella of the co-ranking framework described in [29].”
- Heavy-tailed kernels reveal a finer cluster structure in t-SNE visualisations (2019, background): “Indeed, we found that k-nearest neighbour (KNN) preservation [8] peaked elsewhere: the peak for k = 10 was at α ≈ 1.”
- Type 1 and 2 mixtures of Kullback-Leibler divergences as cost functions in dimensionality reduction based on similarity preservation (2013, background): “Previous work has shown that this immunity stems partly from a property of shift invariance that allows appropriately normalized softmax similarities to mitigate the phenomenon of norm concentration.”
- A new embedding quality assessment method for manifold learning (2011, background): “Lee and Verleysen [34], [35] proposed a general framework, named co-ranking matrix, for rank-based criteria.”