Learning Graph Matching with a Graph-Based Perceptron in a Classification Context

Abstract

Many tasks in computer vision and pattern recognition are formulated as graph matching problems. Despite the NP-hard nature of the problem, fast and accurate approximations have led to significant progress in a wide range of applications. Learning graph matching functions from observed data, however, still remains a challenging issue. This paper presents an effective scheme to parametrize a graph model for object matching in a classification context. For this, we propose a representation based on a parametrized model graph, and optimize it to increase a classification rate. Experimental evaluations on real datasets demonstrate the effectiveness (in terms of accuracy and speed) of our approach against graph classification with hand-crafted cost functions.

Learning error-correcting graph matching with a multiclass neural network

“…The rst development of this idea was in paper (Raveaux et al, 2017). In this paper, we go further in three aspects: (i) Theoretically, we extend our method to multiclass problems.…”

“…(iii) Finally, four datasets are added to the experimental study. The experimental phase shows that the newly proposed approach achieves various improvements compared to the prior approach published in (Raveaux et al, 2017 In this section, the choice of the perceptron algorithm is explained. Then, on a 2-class problem, the key concepts of our graph-based perceptron are presented along with the changes from the original perceptron paradigm (Rosenblatt, 1957).…”

Learning edit cost estimation models for graph edit distance

“…We can separate these approaches in three main groups depending on its goal. By one hand, the first group (Raveaux R., 2017;Bunke, 2005, 2007;Leordeanu et al, 2012) focuses on the graph classification ratio, while the second group (Caetano et al, 2009;Cortés and Serratosa, 2016;Serratosa F., 2011;Cortés and Serratosa, 2015; ?) targets the node-to-node matching accuracy.…”

A Deep Neural Network Architecture to Estimate Node Assignment Costs for the Graph Edit Distance

“…We can divide the learning methods in three main groups depending on the objective function. The first group [7][8][9][10] addresses the recognition ratio for graph classification, while the second group [4,5,11,12] targets the hamming distance. Finally, there is a special case in [13] that does not learn the parameters to estimate the costs but tries to predict if an assignment between nodes is correct or not depending on the values of the costs matrix (the matrix with the costs of each edit operation).…”