Learning from Pairwise Constraints by Similarity Neural Networks
- Day - Time: 21 July 2010, h.11:00
- Place: Area della Ricerca CNR di Pisa - Room: C-29
Speakers
- Lorenzo Sarti (Dipartimento di Ingegneria dell'Informazione - Università di Siena)
Referent
Abstract
We present Similarity Neural Networks (SNNs), a neural network model able to learn a similarity measure for pairs of patterns, exploiting a binary supervision on their similarity/dissimilarity relationships. Pairwise relationships, referred to also as pairwise constraints, generally contain less information than class labels, but, in some contexts, are easier to obtain from human supervisors.
The SNN architecture guarantees the basic properties of a similarity measure (symmetry and non negativity) and, differently from the majority of the metric learning algorithms proposed so far, it can model non-linear relationships among data and can deal with non-transitivity of the similarity criterion. The theoretical properties of SNNs and their application to SemiÂSupervised Clustering will be presented.
In particular, we introduce a novel technique that allows the clustering algorithm to compute the optimal representatives of a data partition by means of Backpropagation on the input layer, biased by a L2 norm regularizer. An extensive set of experimental results will be reported, in order to compare SNNs with state-of-the art similarity learning algorithms. Both on benchmarks and real world data, SNNs and SNN-based clustering show improved performances, assessing the advantage of the proposed neural network approach to similarity measure learning.