Fractional Norms and Quasinorms Do Not Help to Overcome the Curse of Dimensionality
Abstract
:1. Introduction
- The number of attributes for most of real life databases is far from any reasonable intrinsic dimensionality of data;
- The popular estimations of intrinsic dimensionality based on principal components (Kaiser rule and broken stick rule) are very sensitive to irrelevant attributes, while the estimations based on the condition number of the reduced covariance matrix is much more stable as well as the definitions based on separability properties or fractal dimension;
- Usage of functionals with small p does not prevent the concentration of distances;
- A lower value of a distance concentration indicator does not mean better accuracy of the kNN classification.
2. Measure Concentration
- Hubness (popular nearest neighbours) [13].
3. Dimension Estimation
4. Comparison of lp Functionals
4.1. Databases for Comparison
- Data are not time-series.
- Database is formed for the binary classification problem.
- Database does not contain any missing values.
- The number of attributes is less than the number of observations and is greater than 3.
- All predictors are binary or numeric.
- Empty preprocessing means usage data ‘as is’;
- Standardisation means shifting and scaling data to have a zero mean and unit variance;
- Min-max normalization refers to shifting and scaling data in the interval .
4.2. Approaches to Comparison
- The number of databases for which the algorithm is the best [107];
- The number of databases for which the algorithm is the worst [107];
- The number of databases for which the algorithm has performance that is statistically insignificantly different from the best;
- The number of databases for which the algorithm has performance that is statistically insignificantly different from the worst;
- The Wilcoxon signed rank test was used to compare three pairs of metrics.
4.2.1. Proportion Estimation
4.2.2. Friedman Test and Post Hoc Nemenyi Test
4.2.3. Wilcoxon Signed Rank Test
5. Dimension Comparison
- #Attr, PCA-K and PCA-BS;
- PCA-CN and SepD.
- The dimension of the vector space of the dataset is (see Table 4).
- For the dimension defined by the Kaiser rule, PCA-K, the threshold of informativeness is . This means that for all principal components with nonzero eigenvalues, we can take large enough m to ensure that these principal components are “informative” (see Table 4). The significance threshold decreases linearly with increasing m.
- For the dimension defined by the broken stick rule, PCA-BS, we observe initially an increase in the thresholds for the last half of the original principal components, but then the thresholds decrease with an increase in m for all . This means that for all principal components with nonzero eigenvalues, we can take large enough m to ensure that these principal components are “informative” (see Table 4). The thresholds of significance decrease non-linearly with increasing m. This slower than linear thresholds decreasing shows that PCA-BS is less sensitivity to irrelevant attributes than #Attr or PCA-K.
- For the PCA-CN dimension defined by condition number, nothing changes in the described procedure since simultaneous multiplying of all eigenvalues by a nonzero constant does not change the fraction of eigenvalues in the condition (7).
- Adding irrelevant attributes does not change anything for separability dimension, SepD, since the dot product of any two data points in the extended database is the dot products of the corresponding vectors in the original data set multiplied by . This means that described extension of dataset change nothing in the separability inequality (8).
- There are no changes for the fractal dimension FracD, since the described extension of dataset does not change the relative location of data points in space. This means that values will be the same for original and extended datasets.
6. Results of lp Functionals Comparison
7. Discussion
- PCA with Kaiser rule for determining the number of principal components to retain (PCA-K);
- PCA with the broken stick rule for determining the number of principal components to retain (PCA-BS);
- PCA with the condition number criterion for determining the number of principal components to retain (PCA-CN);
- The Fisher separability dimension (SepD);
- The fractal dimension (FracD).
8. Conclusions
Author Contributions
Funding
Conflicts of Interest
Abbreviations
kNN | k nearest neighbours |
RC | relative contrast |
CV | coefficient of variation |
#Attr | the number of attributes or Hamel dimension |
PCA | principle component analysis |
PCA-K | dimension of space according to Kaiser rule for PCA |
PCA-BS | dimension of space according to broken stick rule for PCA |
FVE | fraction of variance explained |
PCA-CN | dimension of space defined by condition number for PCA |
SepD | dimension of space defined according to separability theorem |
FracD | intrinsic dimension defined as fractal dimension |
TNNSC | total number of neighbours of the same class in nearest neighbours of all points |
Se | sensitivity or fraction of correctly recognised cases of positive class |
Sp | specificity or fraction of correctly recognised cases of negative class |
Appendix A. Proof of Propositions
Appendix B. Results for kNN Tests for k = 3, 5, 7
Indicator\p for Functional | 0.01 | 0.1 | 0.5 | 1 | 2 | 4 | 10 | ∞ |
---|---|---|---|---|---|---|---|---|
TNNSC | ||||||||
Empty preprocessing | ||||||||
The best | 3 | 10 | 11 | 1 | 7 | 2 | 0 | 3 |
Insignificantly different from the best | 23 | 28 | 31 | 32 | 32 | 30 | 27 | 25 |
The worst | 20 | 4 | 1 | 1 | 1 | 4 | 4 | 9 |
Insignificantly different from the worst | 32 | 25 | 23 | 23 | 24 | 25 | 26 | 26 |
Standardisation | ||||||||
The best | 1 | 4 | 9 | 7 | 8 | 2 | 3 | 2 |
Insignificantly different from the best | 23 | 28 | 32 | 33 | 32 | 31 | 29 | 26 |
The worst | 17 | 1 | 0 | 0 | 0 | 2 | 4 | 11 |
Insignificantly different from the worst | 34 | 27 | 26 | 24 | 24 | 25 | 25 | 26 |
Min-max normalization | ||||||||
The best | 2 | 4 | 5 | 10 | 8 | 3 | 1 | 6 |
Insignificantly different from the best | 18 | 25 | 31 | 30 | 28 | 28 | 25 | 26 |
The worst | 21 | 3 | 2 | 4 | 2 | 3 | 3 | 8 |
Insignificantly different from the worst | 33 | 24 | 24 | 22 | 21 | 23 | 26 | 27 |
Accuracy | ||||||||
Empty preprocessing | ||||||||
The best | 2 | 11 | 10 | 2 | 8 | 1 | 1 | 2 |
Insignificantly different from the best | 26 | 31 | 33 | 34 | 34 | 33 | 33 | 32 |
The worst | 17 | 4 | 1 | 3 | 1 | 3 | 8 | 7 |
Insignificantly different from the worst | 34 | 28 | 28 | 27 | 26 | 28 | 29 | 29 |
Accuracy | ||||||||
Standardisation | ||||||||
The best | 0 | 4 | 12 | 8 | 8 | 2 | 3 | 2 |
Insignificantly different from the best | 26 | 30 | 34 | 33 | 33 | 33 | 30 | 30 |
The worst | 15 | 2 | 0 | 0 | 2 | 3 | 6 | 8 |
Insignificantly different from the worst | 34 | 29 | 28 | 27 | 27 | 28 | 29 | 29 |
Min-max normalization | ||||||||
The best | 2 | 4 | 9 | 13 | 6 | 4 | 1 | 6 |
Insignificantly different from the best | 27 | 29 | 33 | 32 | 32 | 31 | 31 | 31 |
The worst | 15 | 4 | 5 | 4 | 3 | 4 | 5 | 9 |
Insignificantly different from the worst | 34 | 30 | 27 | 28 | 29 | 29 | 30 | 29 |
Sensitivity plus specificity | ||||||||
Empty preprocessing | ||||||||
The best | 5 | 11 | 9 | 4 | 8 | 2 | 2 | 1 |
The worst | 14 | 2 | 1 | 2 | 0 | 2 | 6 | 9 |
Standardisation | ||||||||
The best | 1 | 7 | 10 | 5 | 8 | 2 | 2 | 3 |
The worst | 12 | 1 | 0 | 1 | 2 | 2 | 7 | 10 |
Min-max normalization | ||||||||
The best | 4 | 6 | 9 | 13 | 6 | 4 | 4 | 2 |
The worst | 14 | 1 | 2 | 3 | 2 | 2 | 5 | 11 |
Indicator\p for functional | 0.01 | 0.1 | 0.5 | 1 | 2 | 4 | 10 | ∞ |
---|---|---|---|---|---|---|---|---|
TNNSC | ||||||||
Empty preprocessing | ||||||||
The best | 2 | 9 | 7 | 3 | 7 | 3 | 2 | 2 |
Insignificantly different from the best | 21 | 28 | 31 | 32 | 32 | 28 | 25 | 22 |
The worst | 20 | 2 | 1 | 1 | 1 | 3 | 2 | 10 |
Insignificantly different from the worst | 31 | 24 | 20 | 22 | 23 | 24 | 25 | 26 |
Standardisation | ||||||||
The best | 0 | 3 | 10 | 9 | 9 | 1 | 2 | 3 |
Insignificantly different from the best | 20 | 28 | 32 | 32 | 31 | 31 | 28 | 25 |
The worst | 20 | 1 | 1 | 2 | 0 | 1 | 2 | 10 |
Insignificantly different from the worst | 34 | 27 | 21 | 20 | 21 | 23 | 26 | 27 |
Min-max normalization | ||||||||
The best | 2 | 4 | 5 | 10 | 8 | 3 | 1 | 6 |
Insignificantly different from the best | 18 | 25 | 31 | 30 | 28 | 28 | 25 | 26 |
The worst | 21 | 3 | 2 | 4 | 2 | 3 | 3 | 8 |
Insignificantly different from the worst | 33 | 24 | 24 | 22 | 21 | 23 | 26 | 27 |
Accuracy | ||||||||
Empty preprocessing | ||||||||
The best | 2 | 13 | 7 | 4 | 10 | 3 | 5 | 3 |
Insignificantly different from the best | 26 | 31 | 32 | 33 | 33 | 33 | 32 | 30 |
The worst | 15 | 2 | 1 | 2 | 4 | 4 | 7 | 10 |
Insignificantly different from the worst | 33 | 29 | 27 | 27 | 26 | 27 | 27 | 28 |
Standardisation | ||||||||
The best | 3 | 11 | 12 | 8 | 12 | 3 | 1 | 3 |
Insignificantly different from the best | 27 | 29 | 33 | 33 | 32 | 32 | 32 | 29 |
The worst | 16 | 2 | 0 | 0 | 0 | 4 | 6 | 8 |
Insignificantly different from the worst | 34 | 29 | 29 | 28 | 28 | 29 | 30 | 30 |
Min-max normalization | ||||||||
The best | 2 | 4 | 9 | 13 | 6 | 4 | 1 | 6 |
Insignificantly different from the best | 27 | 29 | 33 | 32 | 32 | 31 | 31 | 31 |
The worst | 15 | 4 | 5 | 4 | 3 | 4 | 5 | 9 |
Insignificantly different from the worst | 34 | 30 | 27 | 28 | 29 | 29 | 30 | 29 |
Sensitivity plus specificity | ||||||||
Empty preprocessing | ||||||||
The best | 5 | 11 | 7 | 3 | 8 | 4 | 4 | 1 |
The worst | 12 | 0 | 0 | 1 | 3 | 2 | 7 | 11 |
Standardisation | ||||||||
The best | 4 | 11 | 10 | 5 | 11 | 1 | 1 | 2 |
The worst | 13 | 0 | 0 | 0 | 2 | 2 | 8 | 9 |
Min-max normalization | ||||||||
The best | 4 | 6 | 9 | 13 | 6 | 4 | 4 | 2 |
The worst | 14 | 1 | 2 | 3 | 2 | 2 | 5 | 11 |
Indicator\p for functional | 0.01 | 0.1 | 0.5 | 1 | 2 | 4 | 10 | ∞ |
---|---|---|---|---|---|---|---|---|
TNNSC | ||||||||
Empty preprocessing | ||||||||
The best | 0 | 9 | 7 | 5 | 6 | 3 | 1 | 4 |
Insignificantly different from the best | 20 | 28 | 31 | 32 | 28 | 27 | 23 | 23 |
The worst | 22 | 1 | 1 | 1 | 1 | 3 | 2 | 10 |
Insignificantly different from the worst | 32 | 24 | 19 | 21 | 22 | 24 | 25 | 28 |
Standardisation | ||||||||
The best | 0 | 5 | 6 | 12 | 9 | 2 | 1 | 1 |
Insignificantly different from the best | 20 | 26 | 32 | 32 | 31 | 31 | 26 | 25 |
The worst | 20 | 2 | 0 | 0 | 0 | 2 | 2 | 9 |
Insignificantly different from the worst | 34 | 26 | 20 | 20 | 21 | 22 | 25 | 28 |
TNNSC | ||||||||
Min-max normalization | ||||||||
The best | 2 | 4 | 5 | 10 | 8 | 3 | 1 | 6 |
Insignificantly different from the best | 18 | 25 | 31 | 30 | 28 | 28 | 25 | 26 |
The worst | 21 | 3 | 2 | 4 | 2 | 3 | 3 | 8 |
Insignificantly different from the worst | 33 | 24 | 24 | 22 | 21 | 23 | 26 | 27 |
Accuracy | ||||||||
Empty preprocessing | ||||||||
The best | 4 | 12 | 13 | 8 | 8 | 2 | 2 | 5 |
Insignificantly different from the best | 25 | 31 | 33 | 34 | 34 | 34 | 32 | 30 |
The worst | 15 | 2 | 2 | 3 | 4 | 8 | 9 | 9 |
Insignificantly different from the worst | 34 | 29 | 28 | 27 | 27 | 30 | 30 | 29 |
Standardisation | ||||||||
The best | 2 | 4 | 13 | 5 | 10 | 2 | 0 | 2 |
Insignificantly different from the best | 27 | 28 | 34 | 33 | 32 | 32 | 31 | 30 |
The worst | 14 | 4 | 1 | 2 | 1 | 5 | 6 | 8 |
Insignificantly different from the worst | 34 | 30 | 29 | 26 | 28 | 29 | 31 | 31 |
Min-max normalization | ||||||||
The best | 2 | 4 | 9 | 13 | 6 | 4 | 1 | 6 |
Insignificantly different from the best | 27 | 29 | 33 | 32 | 32 | 31 | 31 | 31 |
The worst | 15 | 4 | 5 | 4 | 3 | 4 | 5 | 9 |
Insignificantly different from the worst | 34 | 30 | 27 | 28 | 29 | 29 | 30 | 29 |
Sensitivity plus specificity | ||||||||
Empty preprocessing | ||||||||
The best | 5 | 8 | 9 | 8 | 6 | 3 | 3 | 3 |
The worst | 12 | 1 | 3 | 2 | 4 | 5 | 7 | 11 |
Standardisation | ||||||||
The best | 3 | 6 | 11 | 6 | 7 | 1 | 0 | 2 |
The worst | 11 | 1 | 0 | 0 | 1 | 3 | 4 | 14 |
Min-max normalization | ||||||||
The best | 4 | 6 | 9 | 13 | 6 | 4 | 4 | 2 |
The worst | 14 | 1 | 2 | 3 | 2 | 2 | 5 | 11 |
References
- Bellman, R.E. Adaptive Control Processes: A Guided Tour; Princeton University Press: Princeton, NJ, USA, 1961. [Google Scholar] [CrossRef]
- Bishop, C.M. The curse of dimensionality. In Pattern Recognition and Machine Learning; Springer: New York, NY, USA, 2006; pp. 33–38. [Google Scholar]
- Hastie, T.; Tibshirani, R.; Friedman, J.H. The Elements of Statistical Learning: Data Mining, Inference, and Prediction; Springer: New York, NY, USA, 2009. [Google Scholar] [CrossRef]
- Korn, F.; Pagel, B.U.; Faloutsos, C. On the “dimensionality curse” and the “self-similarity blessing”. IEEE Trans. Knowl. Data Eng. 2001, 13, 96–111. [Google Scholar] [CrossRef] [Green Version]
- Bishop, C.M. Neural Networks for Pattern Recognition; Oxford University Press: New York, NY, USA, 1995. [Google Scholar]
- Gorban, A.N.; Makarov, V.A.; Tyukin, I.Y. The unreasonable effectiveness of small neural ensembles in high-dimensional brain. Phys. Life Rev. 2019, 29, 55–88. [Google Scholar] [CrossRef] [PubMed]
- Billings, S.A.; Wei, H.L.; Balikhin, M.A. Generalized multiscale radial basis function networks. Neural Netw. 2007, 20, 1081–1094. [Google Scholar] [CrossRef] [PubMed]
- Roh, S.B.; Oh, S.K.; Pedrycz, W.; Seo, K.; Fu, Z. Design methodology for radial basis function neural networks classifier based on locally linear reconstruction and conditional fuzzy C-means clustering. Int. J. Approx. Reason. 2019, 106, 228–243. [Google Scholar] [CrossRef]
- Tkachenko, R.; Tkachenko, P.; Izonin, I.; Vitynskyi, P.; Kryvinska, N.; Tsymbal, Y. Committee of the Combined RBF-SGTM Neural-Like Structures for Prediction Tasks. In Mobile Web and Intelligent Information Systems; Awan, I., Younas, M., Ünal, P., Aleksy, M., Eds.; Springer International Publishing: Cham, Switzerland, 2019; pp. 267–277. [Google Scholar] [CrossRef]
- Izonin, I.; Tkachenko, R.; Kryvinska, N.; Gregus, M.; Tkachenko, P.; Vitynskyi, P. Committee of SGTM Neural-Like Structures with RBF kernel for Insurance Cost Prediction Task. In Proceedings of the 2019 IEEE 2nd Ukraine Conference on Electrical and Computer Engineering (UKRCON), Lviv, Ukraine, 2–6 July 2019; pp. 1037–1040. [Google Scholar] [CrossRef]
- Beyer, K.; Goldstein, J.; Ramakrishnan, R.; Shaft, U. When is “nearest neighbor” meaningful? In International Conference on Database Theory; Lecture Notes in Computer Science; Springer: Berlin/Heidelberg, Germany, 1999; pp. 217–235. [Google Scholar] [CrossRef] [Green Version]
- Hinneburg, A.; Aggarwal, C.C.; Keim, D.A. What is the nearest neighbor in high dimensional spaces? In Proceedings of the 26th International Conference on Very Large Databases, Cairo, Egypt, 10–14 September 2000; pp. 506–515. Available online: https://kops.uni-konstanz.de/bitstream/handle/123456789/5849/P506.pdf (accessed on 29 September 2020).
- Radovanović, M.; Nanopoulos, A.; Ivanović, M. Hubs in space: Popular nearest neighbors in high-dimensional data. J. Mach. Learn. Res. 2010, 11, 2487–2531. Available online: http://www.jmlr.org/papers/volume11/radovanovic10a/radovanovic10a.pdf (accessed on 29 September 2020).
- Aggarwal, C.C.; Hinneburg, A.; Keim, D.A. On the surprising behavior of distance metrics in high dimensional space. In International Conference on Database Theory; Springer: Berlin/Heidelberg, Germany, 2001; pp. 420–434. [Google Scholar] [CrossRef] [Green Version]
- Aggarwal, C.C.; Yu, P.S. Outlier Detection for High Dimensional Data. In Proceedings of the 2001 ACM SIGMOD International Conference on Management of Data; Association for Computing Machinery: New York, NY, USA, 2001; pp. 37–46. [Google Scholar] [CrossRef]
- Kainen, P.C. Utilizing geometric anomalies of high dimension: When complexity makes computation easier. In Computer Intensive Methods in Control and Signal Processing; Birkhäuser: Boston, MA, USA, 1997; pp. 283–294. [Google Scholar] [CrossRef]
- Chen, D.; Cao, X.; Wen, F.; Sun, J. Blessing of Dimensionality: High-Dimensional Feature and Its Efficient Compression for Face Verification. In Proceedings of the 2013 IEEE Conference on Computer Vision and Pattern Recognition, Portland, OR, USA, 23–28 June 2013; pp. 3025–3032. [Google Scholar] [CrossRef]
- Gorban, A.N.; Tyukin, I.Y.; Romanenko, I. The blessing of dimensionality: Separation theorems in the thermodynamic limit. IFAC Pap. 2016, 49, 64–69. [Google Scholar] [CrossRef]
- Liu, G.; Liu, Q.; Li, P. Blessing of dimensionality: Recovering mixture data via dictionary pursuit. IEEE Trans. Pattern Anal. Mach. Intell. 2017, 39, 47–60. [Google Scholar] [CrossRef]
- Vershynin, R. High-Dimensional Probability: An Introduction with Applications in Data Science; Cambridge University Press: Cambridge, UK, 2018; Volume 47. [Google Scholar] [CrossRef]
- Gorban, A.N.; Makarov, V.A.; Tyukin, I.Y. High-dimensional brain in a high-dimensional world: Blessing of dimensionality. Entropy 2020, 22, 82. [Google Scholar] [CrossRef] [Green Version]
- Gromov, M. Isoperimetry of waists and concentration of maps. Geom. Funct. Anal. 2003, 13, 178–215. [Google Scholar] [CrossRef] [Green Version]
- Giannopoulos, A.A.; Milman, V.D. Concentration Property on Probability Spaces. Adv. Math. 2000, 156, 77–106. [Google Scholar] [CrossRef] [Green Version]
- Gorban, A.N.; Tyukin, I.Y. Blessing of dimensionality: Mathematical foundations of the statistical physics of data. Philos. Trans. R. Soc. A 2018, 376, 20170237. [Google Scholar] [CrossRef] [Green Version]
- Ledoux, M. The Concentration of Measure Phenomenon; American Mathematical Society: Providence, RI, USA, 2001. [Google Scholar] [CrossRef]
- Donoho, D.L. High-dimensional data analysis: The curses and blessings of dimensionality. In Proceedings of the AMS Conference on Math Challenges of the 21st Century, Los Angeles, CA, USA, 6–12 August 2000; pp. 1–33. Available online: https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.329.3392&rep=rep1&type=pdf (accessed on 29 September 2020).
- Anderson, J.; Belkin, M.; Goyal, N.; Rademacher, L.; Voss, J. The more, the merrier: The blessing of dimensionality for learning large Gaussian mixtures. J. Mach. Learn. Res. Workshop Conf. Proc. 2014, 35, 1135–1164. Available online: http://proceedings.mlr.press/v35/anderson14.pdf (accessed on 29 September 2020).
- Gorban, A.N.; Golubkov, A.; Grechuk, B.; Mirkes, E.M.; Tyukin, I.Y. Correction of AI systems by linear discriminants: Probabilistic foundations. Inform. Sci. 2018, 466, 303–322. [Google Scholar] [CrossRef] [Green Version]
- Gorban, A.N.; Tyukin, I.Y. Stochastic separation theorems. Neural Netw. 2017, 94, 255–259. [Google Scholar] [CrossRef] [Green Version]
- Tyukin, I.Y.; Higham, D.J.; Gorban, A.N. On Adversarial Examples and Stealth Attacks in Artificial Intelligence Systems. arXiv 2020, arXiv:cs.LG/2004.04479. [Google Scholar]
- Brown, A.; Pearcy, C. Introduction to Operator Theory I: Elements of Functional Analysis; Springer: New York, NY, USA, 2012. [Google Scholar] [CrossRef]
- Pearson, K. LIII. On lines and planes of closest fit to systems of points in space. Lond. Edinb. Dublin Philos. Mag. J. Sci. 1901, 2, 559–572. [Google Scholar] [CrossRef] [Green Version]
- Jobson, J.D. Applied Multivariate Data Analysis: Volume II: Categorical and Multivariate Methods; Springer: New York, NY, USA, 1992. [Google Scholar] [CrossRef]
- Guttman, L. Some necessary conditions for common-factor analysis. Psychometrika 1954, 19, 149–161. [Google Scholar] [CrossRef]
- Kaiser, H.F. The application of electronic computers to factor analysis. Educ. Psychol. Meas. 1960, 20, 141–151. [Google Scholar] [CrossRef]
- Jackson, D.A. Stopping rules in principal components analysis: A comparison of heuristical and statistical approaches. Ecology 1993, 74, 2204–2214. [Google Scholar] [CrossRef]
- Fukunaga, K.; Olsen, D.R. An algorithm for finding intrinsic dimensionality of data. IEEE Trans. Comput. 1971, C-20, 176–183. [Google Scholar] [CrossRef]
- Albergante, L.; Bac, J.; Zinovyev, A. Estimating the effective dimension of large biological datasets using Fisher separability analysis. In Proceedings of the 2019 International Joint Conference on Neural Networks (IJCNN), Budapest, Hungary, 14–19 July 2019; pp. 1–8. [Google Scholar] [CrossRef] [Green Version]
- Vicsek, T. Fractal Growth Phenomena; World Scientific Publishing: Singapore, 1992. [Google Scholar]
- Köthe, G. Topological Vector Spaces. Translated by DJH Garling; Springer: New York, NY, USA, 1969. [Google Scholar]
- François, D.; Wertz, V.; Verleysen, M. Non-Euclidean Metrics for Similarity Search in Noisy Datasets. In Proceedings of the ESANN, Bruges, Belgium, 27–29 April 2005; pp. 339–344. Available online: https://www.elen.ucl.ac.be/Proceedings/esann/esannpdf/es2005-116.pdf (accessed on 29 September 2020).
- Francois, D.; Wertz, V.; Verleysen, M. The concentration of fractional distances. IEEE Trans. Knowl. Data Eng. 2007, 19, 873–886. [Google Scholar] [CrossRef]
- Dik, A.; Jebari, K.; Bouroumi, A.; Ettouhami, A. Fractional Metrics for Fuzzy c-Means. Int. J. Comput. Infor. Tech. 2014, 3, 1490–1495. Available online: https://www.ijcit.com/archives/volume3/issue6/Paper030641.pdf (accessed on 29 September 2020).
- Jayaram, B.; Klawonn, F. Can unbounded distance measures mitigate the curse of dimensionality? Int. J. Data Min. Model. Manag. 2012, 4. [Google Scholar] [CrossRef] [Green Version]
- France, S.L.; Carroll, J.D.; Xiong, H. Distance metrics for high dimensional nearest neighborhood recovery: Compression and normalization. Inform. Sci. 2012, 184, 92–110. [Google Scholar] [CrossRef]
- Doherty, K.A.J.; Adams, R.G.; Davey, N. Non-Euclidean norms and data normalisation. In Proceedings of the ESANN 2004, Bruges, Belgium, 28–30 April 2004; pp. 181–186. Available online: https://www.elen.ucl.ac.be/Proceedings/esann/esannpdf/es2004-65.pdf (accessed on 29 September 2020).
- Cormode, G.; Indyk, P.; Koudas, N.; Muthukrishnan, S. Fast mining of massive tabular data via approximate distance computations. In Proceedings of the 18th International Conference on Data Engineering, San Jose, CA, USA, 26 February–1 March 2002; pp. 605–614. [Google Scholar] [CrossRef]
- Datar, M.; Immorlica, N.; Indyk, P.; Mirrokni, V.S. Locality-sensitive hashing scheme based on p-stable distributions. In Proceedings of the Twentieth Annual Symposium on Computational Geometry—SCG ’04; ACM Press: New York, NY, USA, 2004; pp. 253–262. [Google Scholar] [CrossRef] [Green Version]
- Gorban, A.N.; Mirkes, E.M.; Zinovyev, A. Data analysis with arbitrary error measures approximated by piece-wise quadratic PQSQ functions. In Proceedings of the 2018 International Joint Conference on Neural Networks (IJCNN), Rio, Brazil, 8–13 July 2018; pp. 1–8. [Google Scholar] [CrossRef]
- Allen, E.A.; Damaraju, E.; Plis, S.M.; Erhardt, E.B.; Eichele, T.; Calhoun, V.D. Tracking Whole-Brain Connectivity Dynamics in the Resting State. Cereb. Cortex 2014, 24, 663–676. [Google Scholar] [CrossRef]
- Elkan, C. Using the triangle inequality to accelerate k-means. In Proceedings of the 20th International Conference on Machine Learning (ICML-03), Washington, DC, USA, 21–24 August 2003; pp. 147–153. Available online: https://www.aaai.org/Papers/ICML/2003/ICML03-022.pdf (accessed on 29 September 2020).
- Chang, E.; Goh, K.; Sychay, G.; Wu, G. CBSA: Content-based soft annotation for multimodal image retrieval using Bayes point machines. IEEE Trans. Circuits Syst. Video Technol. 2003, 13, 26–38. [Google Scholar] [CrossRef]
- Demartines, P. Analyse de Données Par Réseaux de Neurones Auto-Organisés. Ph.D. Thesis, Grenoble INPG, Grenoble, France, 1994. (In French). [Google Scholar]
- Yianilos, P.N. Excluded middle vantage point forests for nearest neighbor search. In DIMACS Implementation Challenge, ALENEX’99; Citeseer: Princeton, NJ, USA, 1999; Available online: http://pnylab.com/papers/vp2/vp2.pdf (accessed on 29 September 2020).
- Singh, A.; Yadav, A.; Rana, A. K-means with Three different Distance Metrics. Int. J. Comput. Appl. 2013, 67, 13–17. [Google Scholar] [CrossRef]
- Hu, L.Y.; Huang, M.W.; Ke, S.W.; Tsai, C.F. The distance function effect on k-nearest neighbor classification for medical datasets. SpringerPlus 2016, 5, 1304. [Google Scholar] [CrossRef] [Green Version]
- Pestov, V. Is thek-NN classifier in high dimensions affected by the curse of dimensionality? Comput. Math. Appl. 2013, 65, 1427–1437. [Google Scholar] [CrossRef]
- Gorban, A.N.; Allohibi, J.; Mirkes, E.M. Databases and Code for lp Functional Comparison. Available online: https://github.com/Mirkes/Databases-and-code-for-l%5Fp-functional-comparison (accessed on 11 July 2020).
- Mirkes, E.M.; Allohibi, J.; Gorban, A.N. Do Fractional Norms and Quasinorms Help to Overcome the Curse of Dimensionality? In Proceedings of the 2019 International Joint Conference on Neural Networks (IJCNN), Budapest, Hungary, 14–19 July 2019; pp. 1–8. [Google Scholar] [CrossRef]
- Ledesma, R.D.; Valero-Mora, P. Determining the number of factors to retain in EFA: An easy-to-use computer program for carrying out parallel analysis. Pract. Assess. Res. Eval. 2007, 12, 2. [Google Scholar] [CrossRef]
- Cangelosi, R.; Goriely, A. Component retention in principal component analysis with application to cDNA microarray data. Biol. Direct 2007, 2, 2. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Belsley, D.A.; Kuh, E.; Welsch, R.E. Regression Diagnostics: Identifying Influential Data and Sources of Collinearity; John Wiley & Sons: Hoboken, NJ, USA, 2005. [Google Scholar]
- Gorban, A.N.; Kégl, B.; Wunsch II, D.C.; Zinovyev, A.Y. (Eds.) Principal Manifolds for Data Visualization and Dimension Reduction; Lecture Notes in Computational Science and Engineering; Springer: Berlin/Heidelberg, Germany, 2008; Volume 58. [Google Scholar] [CrossRef] [Green Version]
- Gorban, A.N.; Sumner, N.R.; Zinovyev, A.Y. Topological grammars for data approximation. Appl. Math. Lett. 2007, 20, 382–386. [Google Scholar] [CrossRef] [Green Version]
- Gorban, A.N.; Zinovyev, A. Principal manifolds and graphs in practice: From molecular biology to dynamical systems. Int. J. Neural Syst. 2010, 20, 219–232. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Zinovyev, A.; Mirkes, E. Data complexity measured by principal graphs. Comput. Math. Appl. 2013, 65, 1471–1482. [Google Scholar] [CrossRef]
- Carlsson, G. Topology and data. Bull. Amer. Math. Soc. 2009, 46, 255–308. [Google Scholar] [CrossRef] [Green Version]
- Bac, J.; Zinovyev, A. Lizard brain: Tackling locally low-dimensional yet globally complex organization of multi-dimensional datasets. Front. Neurorobot. 2020, 13, 110. [Google Scholar] [CrossRef] [Green Version]
- Albergante, L.; Zinovyev, A.; Bac, J. Data Point Cloud Separability Analysis Based on Linear Fisher Discriminants. Available online: https://github.com/auranic/FisherSeparabilityAnalysis/tree/master/MATLAB (accessed on 11 July 2020).
- You, K. Package Rdimtools. Available online: https://cran.rstudio.com/web/packages/Rdimtools/Rdimtools.pdf (accessed on 11 July 2020).
- Yu, S.; Principe, J.C. Understanding autoencoders with information theoretic concepts. Neural Netw. 2019, 117, 104–123. [Google Scholar] [CrossRef] [Green Version]
- Sammut, C.; Webb, G.I. Encyclopedia of Machine Learning; Springer: New York, NY, USA, 2011. [Google Scholar] [CrossRef]
- Dheeru, D.; Taniskidou, E.K. UCI Machine Learning Repository. Available online: http://archive.ics.uci.edu/ml (accessed on 11 July 2020).
- Blood Transfusion Service Center. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Blood+Transfusion+Service+Center (accessed on 11 July 2020).
- Banknote Authentication. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/banknote+authentication (accessed on 11 July 2020).
- Khozeimeh, F.; Alizadehsani, R.; Roshanzamir, M.; Khosravi, A.; Layegh, P.; Nahavandi, S. An expert system for selecting wart treatment method. Comput. Biol. Med. 2017, 81, 167–175. [Google Scholar] [CrossRef]
- Khozeimeh, F.; Jabbari Azad, F.; Mahboubi Oskouei, Y.; Jafari, M.; Tehranian, S.; Alizadehsani, R.; Layegh, P. Intralesional immunotherapy compared to cryotherapy in the treatment of warts. Int. J. Dermatol. 2017, 56, 474–478. [Google Scholar] [CrossRef]
- Cryotherapy. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Cryotherapy+Dataset+ (accessed on 11 July 2020).
- Vertebral Column. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Vertebral+Column (accessed on 11 July 2020).
- Immunotherapy. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Immunotherapy+Dataset (accessed on 11 July 2020).
- HTRU2. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/HTRU2 (accessed on 11 July 2020).
- Lyon, R.J. HTRU2. Available online: https://figshare.com/articles/dataset/HTRU2/3080389/1 (accessed on 29 September 2020). [CrossRef]
- Lyon, R.J.; Stappers, B.W.; Cooper, S.; Brooke, J.M.; Knowles, J.D. Fifty years of pulsar candidate selection: From simple filters to a new principled real-time classification approach. Mon. Not. R. Astron. Soc. 2016, 459, 1104–1123. [Google Scholar] [CrossRef] [Green Version]
- Indian Liver Patient. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/ILPD+%28Indian+Liver+Patient+Dataset%29 (accessed on 11 July 2020).
- Bhatt, R. Planning-Relax Dataset for Automatic Classification of EEG Signals. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Planning+Relax (accessed on 11 July 2020).
- MAGIC Gamma Telescope. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/MAGIC+Gamma+Telescope (accessed on 11 July 2020).
- EEG Eye State. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/EEG+Eye+State# (accessed on 11 July 2020).
- Climate Model Simulation Crashes. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Climate+Model+Simulation+Crashes (accessed on 11 July 2020).
- Diabetic Retinopathy Debrecen Data Set. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Diabetic+Retinopathy+Debrecen+Data+Set (accessed on 11 July 2020).
- Antal, B.; Hajdu, A. An ensemble-based system for automatic screening of diabetic retinopathy. Knowl. Based Syst. 2014, 60, 20–27. [Google Scholar] [CrossRef] [Green Version]
- SPECTF Heart. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/SPECTF+Heart (accessed on 11 July 2020).
- Breast Cancer Wisconsin (Diagnostic). UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Breast+Cancer+Wisconsin+%28Diagnostic%29 (accessed on 11 July 2020).
- Ionosphere. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Ionosphere (accessed on 11 July 2020).
- Mansouri, K.; Ringsted, T.; Ballabio, D.; Todeschini, R.; Consonni, V. Quantitative structure–activity relationship models for ready biodegradability of chemicals. J. Chem. Inf. Model. 2013, 53, 867–878. [Google Scholar] [CrossRef]
- QSAR biodegradation. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/QSAR+biodegradation (accessed on 11 July 2020).
- MiniBooNE particle identification. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/MiniBooNE+particle+identification (accessed on 11 July 2020).
- Bridge, J.P.; Holden, S.B.; Paulson, L.C. Machine learning for first-order theorem proving. J. Autom. Reason. 2014, 53, 141–172. [Google Scholar] [CrossRef]
- First-order Theorem Proving. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/First-order+theorem+proving (accessed on 11 July 2020).
- Connectionist Bench (Sonar Mines vs. Rocks). UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/connectionist+bench+%28sonar,+mines+vs%2E+rocks%29 (accessed on 11 July 2020).
- Quality Assessment of Digital Colposcopies. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Quality+Assessment+of+Digital+Colposcopies (accessed on 11 July 2020).
- Fernandes, K.; Cardoso, J.S.; Fernandes, J. Transfer learning with partial observability applied to cervical cancer screening. In Iberian Conference on Pattern Recognition and Image Analysis; Springer International Publishing: Cham, Switzerland, 2017; pp. 243–250. [Google Scholar] [CrossRef] [Green Version]
- Huang, G.B.; Ramesh, M.; Berg, T.; Learned-Miller, E. Labeled Faces in the Wild: A Database for Studying Face Recognition in Unconstrained Environments; Technical Report 07-49; University of Massachusetts: Amherst, NJ, USA, 2007; Available online: http://vis-www.cs.umass.edu/lfw/lfw.pdf (accessed on 29 September 2020).
- Musk 1 and 2. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Musk+(Version+1) (accessed on 11 July 2020).
- Guyon, I.; Gunn, S.; Ben-Hur, A.; Dror, G. Result analysis of the NIPS 2003 feature selection challenge. In Advances in Neural Information Processing Systems 17; Saul, L.K., Weiss, Y., Bottou, L., Eds.; MIT Press: Cambridge, UK, 2005; pp. 545–552. Available online: http://papers.nips.cc/paper/2728-result-analysis-of-the-nips-2003-feature-selection-challenge.pdf (accessed on 29 September 2020).
- Madelon. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Madelon (accessed on 11 July 2020).
- Gisette. UCI Machine Learning Repository. Available online: https://archive.ics.uci.edu/ml/datasets/Gisette (accessed on 11 July 2020).
- Demšar, J. Statistical comparisons of classifiers over multiple data sets. J. Mach. Learn. Res. 2006, 7, 1–30. Available online: http://www.jmlr.org/papers/volume7/demsar06a/demsar06a.pdf (accessed on 29 September 2020).
- Friedman, M. A comparison of alternative tests of significance for the problem of m rankings. Ann. Math. Stat. 1940, 11, 86–92. [Google Scholar] [CrossRef]
- Friedman, M. The use of ranks to avoid the assumption of normality implicit in the analysis of variance. J. Am. Stat. Assoc. 1937, 32, 675–701. [Google Scholar] [CrossRef]
- Nemenyi, P. Distribution-free multiple comparisons. In Biometrics; International Biometric Society: Washington, DC, USA, 1962; Volume 18, p. 263. [Google Scholar]
- Altman, D.; Machin, D.; Bryant, T.; Gardner, M. (Eds.) Statistics with Confidence: Confidence Intervals and Statistical Guidelines, 2nd ed.; John Wiley & Sons: Hoboken, NJ, USA, 2013. [Google Scholar]
- Biau, D.J.; Kernéis, S.; Porcher, R. Statistics in brief: The importance of sample size in the planning and interpretation of medical research. Clin. Orthop. Relat. R. 2008, 466, 2282–2288. [Google Scholar] [CrossRef] [Green Version]
- Kadam, P.; Bhalerao, S. Sample size calculation. Int. J. Ayurveda Res. 2010, 1, 55–57. [Google Scholar] [CrossRef] [Green Version]
- Bonferroni, C.E. Il calcolo delle assicurazioni su gruppi di teste. In Studi in Onore del Professore Salvatore Ortu Carboni; Tip. del Senato: Roma, Italy, 1935; pp. 13–60. (In Italian) [Google Scholar]
- Wilcoxon, F. Individual comparisons by ranking methods. Biom. Bull. 1945, 1, 80–83. [Google Scholar] [CrossRef]
- Wilcoxon Signed Rank Test. Available online: https://www.mathworks.com/help/stats/signrank.html (accessed on 11 July 2020).
- Talagrand, M. Concentration of measure and isoperimetric inequalities in product spaces. Publ. Math. 1995, 81, 73–205. [Google Scholar] [CrossRef] [Green Version]
- Guédon, O.; Milman, E. Interpolating thin-shell and sharp large-deviation estimates for isotropic log-concave measures. Geom. Funct. Anal. 2011, 21, 1043. [Google Scholar] [CrossRef]
- Tibshirani, R. Regression shrinkage and selection via the lasso. J. R. Stat. Soc. B 1996, 58, 267–288. [Google Scholar] [CrossRef]
Dim | for # of Points | |||
---|---|---|---|---|
10 [14] | 10 | 20 | 100 | |
1 | 0 | 0 | 0 | 0 |
2 | 0.850 | 0.850 | 0.960 | 1.00 |
3 | 0.887 | 0.930 | 0.996 | 1.00 |
4 | 0.913 | 0.973 | 0.996 | 1.00 |
10 | 0.956 | 0.994 | 1.00 | 1.00 |
15 | 0.961 | 1.000 | 1.00 | 1.00 |
20 | 0.971 | 0.999 | 1.00 | 1.00 |
100 | 0.982 | 1.000 | 1.00 | 1.00 |
Name | Source | #Attr. | Cases | PCA-K | PCA-BS | PCA-CN | SepD | FracD |
---|---|---|---|---|---|---|---|---|
Blood | [74] | 4 | 748 | 2 | 2 | 3 | 2.4 | 1.6 |
Banknote authentication | [75] | 4 | 1372 | 2 | 2 | 3 | 2.6 | 1.9 |
Cryotherapy | [76,77,78] | 6 | 90 | 3 | 0 | 6 | 4.1 | 2.5 |
Vertebral Column | [79] | 6 | 310 | 2 | 1 | 5 | 4.4 | 2.3 |
Immunotherapy | [76,77,80] | 7 | 90 | 3 | 0 | 7 | 5.1 | 3.2 |
HTRU2 | [81,82,83] | 8 | 17,898 | 2 | 2 | 4 | 3.06 | 2.4 |
ILPD (Indian LiverPatient Dataset) | [84] | 10 | 579 | 4 | 0 | 7 | 4.3 | 2.1 |
Planning Relax | [85] | 10 | 182 | 4 | 0 | 6 | 6.09 | 3.6 |
MAGIC Gamma Telescope | [86] | 10 | 19,020 | 3 | 1 | 6 | 4.6 | 2.9 |
EEG Eye State | [87] | 14 | 14,980 | 4 | 4 | 5 | 2.1 | 1.2 |
Climate Model SimulationCrashes | [88] | 18 | 540 | 10 | 0 | 18 | 16.8 | 21.7 |
Diabetic Retinopathy Debrecen | [89,90] | 19 | 1151 | 5 | 3 | 8 | 4.3 | 2.3 |
SPECT Heart | [91] | 22 | 267 | 7 | 3 | 12 | 4.9 | 11.5 |
Breast Cancer | [92] | 30 | 569 | 6 | 3 | 5 | 4.3 | 3.5 |
Ionosphere | [93] | 34 | 351 | 8 | 4 | 9 | 3.9 | 3.5 |
QSAR biodegradation | [94,95] | 41 | 1055 | 11 | 6 | 15 | 5.4 | 3.1 |
SPECTF Heart | [91] | 44 | 267 | 10 | 3 | 6 | 5.6 | 7 |
MiniBooNE particleidentification | [96] | 50 | 130,064 | 4 | 1 | 1 | 0.5 | 2.7 |
First-order theorem proving(6 tasks) | [97,98] | 51 | 6118 | 13 | 7 | 9 | 3.4 | 2.04 |
Connectionist Bench (Sonar) | [99] | 60 | 208 | 13 | 6 | 11 | 6.1 | 5.5 |
Quality Assessment ofDigital Colposcopies (7 tasks) | [100,101] | 62 | 287 | 11 | 6 | 9 | 5.6 | 4.7 |
LFW | [102] | 128 | 13,233 | 51 | 55 | 57 | 13.8 | 19.3 |
Musk 1 | [103] | 166 | 476 | 23 | 9 | 7 | 4.1 | 4.4 |
Musk 2 | [103] | 166 | 6598 | 25 | 13 | 6 | 4.1 | 7.8 |
Madelon | [104,105] | 500 | 2600 | 224 | 0 | 362 | 436.3 | 13.5 |
Gisette | [104,106] | 5000 | 7000 | 1465 | 133 | 25 | 10.2 | 2.04 |
Dimension | #Attr | PCA-K | PCA-BS | PCA-CN | SepD | FracD |
---|---|---|---|---|---|---|
#Attr | 1.000 | 0.998 | 0.923 | 0.098 | 0.065 | −0.081 |
PCA-K | 0.998 | 1.000 | 0.917 | 0.154 | 0.119 | −0.057 |
PCA-BS | 0.923 | 0.917 | 1.000 | 0.018 | −0.058 | 0.075 |
PCA-CN | 0.098 | 0.154 | 0.018 | 1.000 | 0.992 | 0.405 |
SepD | 0.065 | 0.119 | −0.058 | 0.992 | 1.000 | 0.343 |
FracD | −0.081 | −0.057 | 0.075 | 0.405 | 0.343 | 1.000 |
Musk | Gizette | |||||
---|---|---|---|---|---|---|
m | #Attr | PCA-K | PCA-BS | #Attr | PCA-K | PCA-BS |
0 | 166 | 23 | 9 | 4971 | 1456 | 131 |
1 | 332 | 34 | 16 | 9942 | 2320 | 1565 |
2 | 498 | 40 | 23 | 14,913 | 2721 | 1976 |
3 | 664 | 45 | 28 | 19,884 | 2959 | 2217 |
4 | 830 | 49 | 32 | 24,855 | 3122 | 2389 |
5 | 996 | 53 | 33 | 29,826 | 3242 | 2523 |
10 | 1826 | 63 | 39 | 54,681 | 3594 | 2909 |
50 | 8466 | 94 | 62 | 253,521 | 4328 | 3641 |
100 | 16,766 | 109 | 73 | 502,071 | 4567 | 3926 |
500 | 83,166 | 139 | 102 | 2,490,471 | 4847 | 4491 |
1000 | 166,166 | 150 | 115 | 4,975,971 | 4863 | 4664 |
5000 | 830,166 | 163 | 141 | 24,859,971 | 4865 | 4852 |
10,000 | 1,660,166 | 166 | 151 | 49,714,971 | 4866 | 4863 |
Indicator\p for Functional | 0.01 | 0.1 | 0.5 | 1 | 2 | 4 | 10 | ∞ |
---|---|---|---|---|---|---|---|---|
TNNSC | ||||||||
Empty preprocessing | ||||||||
The best | 2 | 11 | 5 | 10 | 7 | 1 | 1 | 1 |
Insignificantly different from the best | 17 | 26 | 31 | 31 | 31 | 30 | 23 | 22 |
The worst | 19 | 0 | 1 | 0 | 1 | 3 | 4 | 8 |
Insignificantly different from the worst | 34 | 23 | 17 | 19 | 21 | 21 | 25 | 29 |
Standardisation | ||||||||
The best | 0 | 5 | 10 | 11 | 6 | 2 | 1 | 1 |
Insignificantly different from the best | 19 | 26 | 33 | 32 | 31 | 30 | 25 | 24 |
The worst | 18 | 2 | 0 | 0 | 1 | 2 | 4 | 10 |
Insignificantly different from the worst | 35 | 24 | 20 | 19 | 20 | 21 | 25 | 28 |
Min-max normalization | ||||||||
The best | 1 | 5 | 10 | 13 | 4 | 6 | 1 | 3 |
Insignificantly different from the best | 19 | 26 | 32 | 31 | 30 | 29 | 26 | 26 |
The worst | 23 | 4 | 2 | 2 | 3 | 3 | 4 | 7 |
Insignificantly different from the worst | 36 | 24 | 22 | 21 | 22 | 22 | 26 | 26 |
Accuracy | ||||||||
Empty preprocessing | ||||||||
The best | 3 | 9 | 9 | 15 | 6 | 5 | 1 | 2 |
Insignificantly different from the best | 29 | 31 | 34 | 35 | 35 | 35 | 33 | 30 |
The worst | 13 | 3 | 1 | 2 | 4 | 4 | 9 | 14 |
Insignificantly different from the worst | 35 | 32 | 28 | 28 | 29 | 29 | 30 | 31 |
Standardisation | ||||||||
The best | 2 | 5 | 12 | 18 | 7 | 3 | 1 | 1 |
Insignificantly different from the best | 30 | 31 | 34 | 34 | 33 | 31 | 32 | 30 |
The worst | 13 | 4 | 0 | 0 | 2 | 6 | 7 | 13 |
Insignificantly different from the worst | 35 | 32 | 29 | 29 | 30 | 31 | 33 | 33 |
Accuracy | ||||||||
Min-max normalization | ||||||||
The best | 2 | 7 | 15 | 8 | 8 | 3 | 3 | 6 |
Insignificantly different from the best | 30 | 31 | 34 | 33 | 33 | 32 | 31 | 32 |
The worst | 18 | 6 | 3 | 4 | 5 | 9 | 8 | 8 |
Insignificantly different from the worst | 36 | 33 | 31 | 31 | 31 | 32 | 33 | 32 |
Sensitivity plus specificity | ||||||||
Empty preprocessing | ||||||||
The best | 4 | 8 | 7 | 12 | 7 | 5 | 1 | 1 |
The worst | 14 | 2 | 1 | 1 | 3 | 5 | 8 | 12 |
Standardisation | ||||||||
The best | 4 | 7 | 8 | 15 | 7 | 2 | 1 | 0 |
The worst | 13 | 3 | 0 | 0 | 2 | 5 | 4 | 15 |
Min-max normalization | ||||||||
The best | 5 | 8 | 13 | 6 | 9 | 3 | 4 | 5 |
The worst | 15 | 4 | 2 | 3 | 3 | 7 | 8 | 13 |
Preprocessing | Quality Indicator | Friedman’s p-Value | The Best | Set of Insignificantly Different from the Best | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
p | 0.01 | 0.1 | 0.5 | 1 | 2 | 4 | 10 | ∞ | ||||
Empty | TNNSC | <0.0001 | 1 | 6.2639 | X | X | X | X | X | |||
Accuracy | <0.0001 | 1 | 6.2639 | X | X | X | X | |||||
Se+Sp | <0.0001 | 0.5 | 6.0556 | X | X | X | X | |||||
Standardisation | TNNSC | <0.0001 | 1 | 6.6944 | X | X | X | |||||
Accuracy | <0.0001 | 1 | 6.8056 | X | X | X | ||||||
Se+Sp | <0.0001 | 1 | 6.4722 | X | X | X | X | |||||
Min-max normalization | TNNSC | <0.0001 | 1 | 6.4722 | X | X | X | X | ||||
Accuracy | <0.0001 | 0.5 | 6.0000 | X | X | X | X | |||||
Se+Sp | <0.0001 | 0.5 | 6.0000 | X | X | X | X |
Preprocessing | Quality Indicator | Friedman’s p-Value | The Best | Set of Insignificantly Different from the Best | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
p | 0.01 | 0.1 | 0.5 | 1 | 2 | 4 | 10 | ∞ | ||||
Empty | TNNSC | <0.0001 | 0.5 | 6.0294 | X | X | X | X | X | |||
Accuracy | <0.0001 | 0.5 | 5.9265 | X | X | X | X | |||||
Se+Sp | <0.0001 | 0.5 | 5.7353 | X | X | X | X | X | ||||
Standardisation | TNNSC | <0.0001 | 1 | 6.2941 | X | X | X | X | ||||
Accuracy | <0.0001 | 0.5 | 6.3235 | X | X | X | X | |||||
Se+Sp | <0.0001 | 0.5 | 6.1324 | X | X | X | X | X | ||||
Min-max normalization | TNNSC | <0.0001 | 2 | 6.0588 | X | X | X | X | ||||
Accuracy | <0.0001 | 1 | 6.0000 | X | X | X | X | X | ||||
Se+Sp | <0.0001 | 1 | 6.0147 | X | X | X | X | X |
Preprocessing | Quality Indicator | Friedman’s p-Value | The Best | Set of Insignificantly Different from the Best | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
p | 0.01 | 0.1 | 0.5 | 1 | 2 | 4 | 10 | ∞ | ||||
Empty | TNNSC | <0.0001 | 0.5 | 5.9118 | X | X | X | X | X | |||
Accuracy | <0.0001 | 0.5 | 5.8971 | X | X | X | X | X | ||||
Se+Sp | <0.0001 | 0.5 | 5.9853 | X | X | X | X | |||||
Standardisation | TNNSC | <0.0001 | 1 | 6.1471 | X | X | X | X | ||||
Accuracy | <0.0001 | 0.5 | 6.1618 | X | X | X | X | |||||
Se+Sp | <0.0001 | 1 | 6.1765 | X | X | X | X | |||||
Min-max normalization | TNNSC | <0.0001 | 2 | 6.0588 | X | X | X | X | ||||
Accuracy | <0.0001 | 1 | 6.0000 | X | X | X | X | X | ||||
Se+Sp | <0.0001 | 1 | 6.0147 | X | X | X | X | X |
Preprocessing | Quality Indicator | Friedman’s p-Value | The Best | Set of Insignificantly Different from the Best | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
p | 0.01 | 0.1 | 0.5 | 1 | 2 | 4 | 10 | ∞ | ||||
Empty | TNNSC | <0.0001 | 1 | 6.1618 | X | X | X | X | X | |||
Accuracy | <0.0001 | 0.5 | 5.8971 | X | X | X | X | |||||
Se+Sp | <0.0001 | 1 | 5.8971 | X | X | X | X | |||||
Standardisation | TNNSC | <0.0001 | 1 | 6.5147 | X | X | X | X | ||||
Accuracy | <0.0001 | 0.5 | 6.3971 | X | X | X | ||||||
Se+Sp | <0.0001 | 0.5 | 6.1176 | X | X | X | X | |||||
Min-max normalization | TNNSC | <0.0001 | 2 | 6.0588 | X | X | X | X | ||||
Accuracy | <0.0001 | 1 | 6.0000 | X | X | X | X | X | ||||
Se+Sp | <0.0001 | 1 | 6.0147 | X | X | X | X | X |
Preprocessing | Quality Indicator | p-Value for and | ||
---|---|---|---|---|
0.5 & 1 | 0.5 & 2 | 1 & 2 | ||
Empty | TNNSC | 0.6348 | 0.3418 | 0.0469 |
Accuracy | 0.9181 | 0.0657 | 0.0064 | |
Se+Sp | 0.8517 | 0.0306 | 0.0022 | |
Standardised | TNNSC | 0.3098 | 0.1275 | 0.0014 |
Accuracy | 0.6680 | 0.0202 | 0.0017 | |
Se+Sp | 0.8793 | 0.0064 | 0.0011 | |
Min-max normalization | TNNSC | 0.7364 | 0.0350 | 0.0056 |
Accuracy | 0.1525 | 0.0218 | 0.2002 | |
Se+Sp | 0.1169 | 0.0129 | 0.3042 |
Quality Indicator | p of Function | p-Value for Pair of Preprocessings | ||
---|---|---|---|---|
E & S | E & M | S & M | ||
TNNSC | 0.5 | 0.5732 | 0.8382 | 0.6151 |
1 | 0.9199 | 0.5283 | 0.1792 | |
2 | 0.9039 | 0.3832 | 0.1418 | |
Accuracy | 0.5 | 0.8446 | 0.5128 | 0.3217 |
1 | 0.8788 | 0.0126 | 0.0091 | |
2 | 0.5327 | 0.3127 | 0.3436 | |
Se+Sp | 0.5 | 0.6165 | 0.2628 | 0.0644 |
1 | 0.5862 | 0.0054 | 0.0067 | |
2 | 0.6292 | 0.3341 | 0.4780 |
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
Share and Cite
Mirkes, E.M.; Allohibi, J.; Gorban, A. Fractional Norms and Quasinorms Do Not Help to Overcome the Curse of Dimensionality. Entropy 2020, 22, 1105. https://doi.org/10.3390/e22101105
Mirkes EM, Allohibi J, Gorban A. Fractional Norms and Quasinorms Do Not Help to Overcome the Curse of Dimensionality. Entropy. 2020; 22(10):1105. https://doi.org/10.3390/e22101105
Chicago/Turabian StyleMirkes, Evgeny M., Jeza Allohibi, and Alexander Gorban. 2020. "Fractional Norms and Quasinorms Do Not Help to Overcome the Curse of Dimensionality" Entropy 22, no. 10: 1105. https://doi.org/10.3390/e22101105
APA StyleMirkes, E. M., Allohibi, J., & Gorban, A. (2020). Fractional Norms and Quasinorms Do Not Help to Overcome the Curse of Dimensionality. Entropy, 22(10), 1105. https://doi.org/10.3390/e22101105