Information Technology Reference
In-Depth Information
iterative UNN strategies are efficient methods to embed high-dimensional data into
fixed one-dimensional latent space taking linear time. UNN achieves lower DSREs,
but UNN g is slightly faster because of the multiplicative constants of UNN. We con-
centrated on the employment of the -insensitive loss, and its influence on the DSRE. It
could be observed that both iterative UNN regression strategies could benefit from the
-insensitive loss, in particular the iterative variant UNN g could be improved employ-
ing a loss with > 0 . Obviously, local optima can be avoided.
We have introduced two algorithmic variants that allow the efficient embedding of
incomplete data. The results have shown that the embeddings are better in case of repair-
and-embed, as it is obviously difficult to determine the embedding of the data with in-
complete patterns. From the perspective of imputation, first repairing incomplete data is
a straightforward approach. In contrast, first embedding data at the location with lowest
DSRE, and then repairing the entries employing the neighbors is an approach that makes
use of the intrinsic structure UNN regression assumes for imputation leading to compar-
atively good pattern reconstructions, but worse embeddings than repair-and-embed.
References
1. Sdss 2011, sloan digital sky survey (2011), http://www.sdss.org
2. Baillard, A., Bertin, E., de Lapparent, V., Fouque, P., Arnouts, S., Mellier, Y., Pell o, R.,
Leborgne, J.-F., Prugniel, P., Markarov, D., Makarova, L., McCracken, H.J., Bijaoui, A.,
Tasca, L.: Galaxy morphology without classification: Self-organizing maps, 532, A74,
1103.5734 (2011)
3. Bhatia, N., Vandana: Survey of nearest neighbor techniques. CoRR, abs/1007.0085 (2010)
4. Bishop, C.M.: Pattern Recognition and Machine Learning (Information Science and Statis-
tics). Springer (2007)
5. Carreira-Perpinan, M. A., Lu, Z.: Parametric dimensionality reduction by unsupervised re-
gression. In: Computer Vision and Pattern Recognition (CVPR), pp. 1895-1902 (2010)
6. Chechik, G., Heitz, G., Elidan, G., Abbeel, P., Koller, D.: Max-margin classification of data
with absent features. Journal of Machine Learning Research 9, 1-21 (2008)
7. Dick, U., Haider, P., Scheffer, T.: Learning from incomplete data with infinite imputations.
In: International Conference on Machine Learning (ICML), pp. 232-239 (2008)
8. Fix, E., Hodges, J.: Discriminatory analysis, nonparametric discrimination: Consistency
properties, vol. 4 (1951)
9. Ghahramani, Z., Jordan, M.I.: Supervised learning from incomplete data via an em approach.
In: Advances in Neuronal Information Processing (NIPS), pp. 120-127 (1993)
10. Gieseke, F., Polsterer, K.L., Thom, A., Zinn, P., Bomanns, D., Dettmar, R.-J., Kramer, O.,
Vahrenhold, J.: Detecting quasars in large-scale astronomical surveys. In: ICMLA, pp. 352-
357 (2010)
11. Hastie, T., Tibshirani, R., Friedman, J.: The Elements of Statistical Learning. Springer, Berlin
(2009)
12. Hastie, Y., Stuetzle, W.: Principal curves. Journal of the American Statistical Associa-
tion 85(406), 502-516 (1989)
13. Hull, J.: A database for handwritten text recognition research. IEEE PAMI 5(16), 550-554
(1994)
14. Jolliffe, I.: Principal component analysis. Springer series in statistics. Springer, New York
(1986)
 
Search WWH ::




Custom Search