Sign constrained rectifier networks with applications to pattern decompositions
dc.contributor.author | An, Senjian | |
dc.contributor.author | Ke, Q. | |
dc.contributor.author | Bennamoun, M. | |
dc.contributor.author | Boussaid, F. | |
dc.contributor.author | Sohel, F. | |
dc.date.accessioned | 2018-08-08T04:42:43Z | |
dc.date.available | 2018-08-08T04:42:43Z | |
dc.date.created | 2018-08-08T03:50:34Z | |
dc.date.issued | 2015 | |
dc.identifier.citation | An, S. and Ke, Q. and Bennamoun, M. and Boussaid, F. and Sohel, F. 2015. Sign constrained rectifier networks with applications to pattern decompositions, pp. 546-559. | |
dc.identifier.uri | http://hdl.handle.net/20.500.11937/69894 | |
dc.identifier.doi | 10.1007/978-3-319-23528-8_34 | |
dc.description.abstract |
© Springer International Publishing Switzerland 2015. In this paper we introduce sign constrained rectifier networks (SCRN), demonstrate their universal classification power and illustrate their applications to pattern decompositions.We prove that the proposed two-hidden-layer SCRN, with sign constraints on the weights of the output layer and on those of the top hidden layer, are capable of separating any two disjoint pattern sets. Furthermore, a two-hidden-layer SCRN of a pair of disjoint pattern sets can be used to decompose one of the pattern sets into several subsets so that each subset is convexly separable from the entire other pattern set; and a single-hidden-layer SCRN of a pair of convexly separable pattern sets can be used to decompose one of the pattern sets into several subsets so that each subset is linearly separable from the entire other pattern set. SCRN can thus be used to learn the pattern structures from the decomposed subsets of patterns and to analyse the discriminant factors of different patterns from the linear classifiers of the linearly separable subsets in the decompositions. With such pattern decompositions exhibiting convex separability or linear separability, users can also analyse the complexity of the classification problem, remove the outliers and the non-crucial points to improve the training of the traditional unconstrained rectifier networks in terms of both performance and efficiency. | |
dc.title | Sign constrained rectifier networks with applications to pattern decompositions | |
dc.type | Conference Paper | |
dcterms.source.volume | 9284 | |
dcterms.source.startPage | 546 | |
dcterms.source.endPage | 559 | |
dcterms.source.title | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | |
dcterms.source.series | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | |
dcterms.source.isbn | 9783319235271 | |
curtin.department | School of Electrical Engineering, Computing and Mathematical Science (EECMS) | |
curtin.accessStatus | Fulltext not available |
Files in this item
Files | Size | Format | View |
---|---|---|---|
There are no files associated with this item. |