An Enhanced Filtering-Based Information Granulation Procedure for Graph Embedding and Classification

Abstract

Granular Computing is a powerful information processing paradigm for synthesizing advanced pattern recognition systems in non-conventional domains. In this paper, a novel procedure for the automatic synthesis of suitable information granules is proposed. The procedure leverages a joint sensitivity-vs-specificity score that accounts the meaningfulness of candidate information granules for each class considered in the classification problem at hand. Only statistically relevant granules are retained for a graph embedding procedure towards a geometric space, in which standard classification systems can be used without alterations. Performance tests have been carried out by considering open access datasets of fully labelled graphs with arbitrarily complex nodes and/or edges attributes that, by definition, must rely on inexact graph matching procedures to quantify dissimilarities. Two variants of the procedure are investigated: a standard variant, which aims at automatically finding suitable information granules for solving the classification problem as a whole, and a class-specific metric learning variant, in which the optimization procedure is performed in a class-aware fashion. In the latter case, each class will have its own set of information granules, along with the corresponding parameters defining distinct instances of the dissimilarity measure. Computational results show that the proposed algorithm is able to outperform the vast majority of current approaches for graph classification, while at the same time returning a grey-box model, interpretable by field-experts.

Relaxed Dissimilarity-based Symbolic Histogram Variants for Granular Graph Embedding

“…Full details on nBMF, along with detailed pseudocodes, can be found in (Baldini. et al, 2019;Martino and Rizzi, 2021). Finally, each cluster proves its validity by comparing its own quality F (C) with a threshold τ ∈ [0, 1].…”

“…GREC is the only dataset for which the dissimilarity measures between nodes and edges are parametric themselves: such values populate γ γ γ which shall be optimized, as described in Section 2.4. Full details on the dissimilarity measures for AIDS, Letter and GREC can be found in (Martino and Rizzi, 2021) and (Baldini et al, 2021). For Mutagenicity, as instead, the dissimilarity measures between nodes and edges are plain non-parametric simple matching between their respective labels.…”

“…• several embedding techniques (Riesen and Bunke, 2009b;Gibert et al, 2011), including GrC-based (Martino et al, 2019a;Baldini. et al, 2019;Martino and Rizzi, 2021;Baldini et al, 2021) and neural-based ones (Bacciu et al, 2018;Martineau et al, 2020).…”

“…Clearly, GRALG also whether equipped with 'relaxed' symbolic histograms, is able to reach state-ofthe-art performances for 4 out of 6 datasets while, at the same time, is able to return an interpretable model: the same peculiarity only holds in other GrCbased pattern recognition systems (namely (Martino et al, 2019a;Baldini. et al, 2019;Martino and Rizzi, 2021;Baldini et al, 2021)), whilst it is well-known that other learning paradigms, notably those based on artificial neural networks (Xu et al, 2019), lack any interpretation.…”

Graph-Based Multi-Label Classification for WiFi Network Traffic Analysis

“…The difference between the order of the two graphs accounts for the nodes' deletion and insertion, whereas the existence of induced edges in only one of the two graphs yields an edge deletion or insertion. Full mathematical details on nBMF can be found in [25].…”