Triangular Similarity Metric Learning: a Siamese Architecture Approach. ( L'apprentissage de similarité triangulaire en utilisant des réseaux siamois)
نویسنده
چکیده
In many machine learning and pattern recognition tasks, there is always a need for appropriate metric functions to measure pairwise distance or similarity between data, where a metric function is a function that defines a distance or similarity between each pair of elements of a set. In this thesis, we propose Triangular Similarity Metric Learning (TSML) for automatically specifying a metric from data. A TSML system is loaded in a siamese architecture which consists of two identical sub-systems sharing the same set of parameters. Each sub-system processes a single data sample and thus the whole system receives a pair of data as the input. The TSML system includes a cost function parameterizing the pairwise relationship between data and a mapping function allowing the system to learn high-level features from the training data. In terms of the cost function, we first propose the Triangular Similarity, a novel similarity metric which is equivalent to the well-known Cosine Similarity in measuring a data pair. Based on a simplified version of the Triangular Similarity, we further develop the triangular loss function in order to perform metric learning, i.e. to increase the similarity between two vectors in the same class and to decrease the similarity between two vectors of different classes. Compared with other distance or similarity metrics, the triangular loss and its gradient naturally offer us an intuitive and interesting geometrical interpretation of the metric learning objective. In terms of the mapping function, we introduce three different options: a linear mapping realized by a simple transformation matrix, a nonlinear mapping realized by Multi-layer Perceptrons (MLP) and a deep nonlinear mapping realized by Convolutional Neural Networks (CNN). With these mapping functions, we present three different TSML systems for various applications, namely, pairwise verification, object identification, dimensionality reduction and data visualization. For each application, we carry out extensive experiments on popular benchmarks and datasets to demonstrate the effectiveness of the proposed systems.
منابع مشابه
Scale-Invariance of Support Vector Machines based on the Triangular Kernel
This report focuses on the scale-invariance and the good performances of Support Vector Machines based on the triangular kernel. After a mathematical analysis of the scale-invariance of learning with that kernel, we illustrate its behavior with a simple 2D classi cation problem and compare its performances to those of a Gaussian kernel on face detection and handwritten character recognition Key...
متن کاملKWSim: Concepts Similarity Measure
The comparison of manually annotated medical images can be done using the comparison of keywords in a lexical way or using the existing medical thesauri to calculate semantic similarity. In this paper, first we introduce the KWSim measure, a fully automated technique of measuring semantic similarity by mapping concepts(keywords) to different medical thesauri and examining the “is-a” relation ty...
متن کاملSupervised Metric Learning with Generalization Guarantees
In recent years, the crucial importance of metrics in machine learning algorithms has led to anincreasing interest in optimizing distance and similarity functions using knowledge from training data to makethem suitable for the problem at hand. This area of research is known as metric learning. Existing methodstypically aim at optimizing the parameters of a given metric with respect ...
متن کاملMesure de similarité pondérée dans l'espace 2D: Application à la reconnaissance de visages
RÉSUMÉ. Cet article propose une nouvelle mesure de similarité pondérée basée sur des matrices pour la classification et la reconnaissance de visages. Le calcul de distances s’effectue entre deux matrices caractéristiques obtenues par deux méthodes bidimensionnelles à savoir l'Analyse en Composantes Principales (ACP2D) et l'Analyse Discriminante Linéaire (ADL2D). Les poids de pondération utilisé...
متن کاملColoration de nombre de Grundy pour les graphes triangulés
Notre travail s’intègre dans la problématique générale de la stabilité du réseau ad hoc. Plusieurs, travaux ont attaqué ce problème. Parmi ces travaux, on trouve la modélisation du réseau ad hoc sous forme d’un graphe (les machines correspondent aux nœuds, les arrêtes correspondent aux liens entre les machines). Donc le problème de stabilité du réseau ad hoc qui correspond à un problème d’alloc...
متن کامل