In this thesis, we mainly consider two closely related classical and general problems in data mining and machine learning: (1) developing new similarity measures between graphs and between nodes of a graph, and (2) developing new density measures on nodes of a graph. In the first part of this thesis, we introduce three new graph kernels (which can be seen as similarity measures between graphs) and assess them on a graph classification task: using molecules from various origins as graphs, we try to classify them according to one of their property (mutagenicity,...). We then study the problem of quantifying the similarity between nodes of graphs. We define three new measures by assigning the forests present in the graph a probability of being...
Measures of similarity play a subtle but important role in a large number of disciplines. For exampl...
Graphs are ubiquitous in many fields of research ranging from sociology to biology. A graph is a ver...
How can we find a good graph clustering of a real-world network, that allows insight into its underl...
Abstract—This work introduces a novel nonparametric density index defined on graphs, the Sum-over-Fo...
This thesis is organized in two parts : the first part focuses on measures of similarity (or proximi...
This work introduces a novel way to identify dense regions in a graph based on a mode-seeking cluste...
Information networks, such as biological or social networks, contain groups of related entities, whi...
Information networks, such as biological or social networks, contain groups of related entities, whi...
Graph clustering methods such as spectral clustering are defined for general weighted graphs. In mac...
In this thesis, we propose many developments in the context of Structural Similarity. We address bot...
Graph clustering methods such as spectral clustering are defined for general weighted graphs. In mac...
We study clustering over multiple graphs- each encoding a distinct set of similarity relationships (...
Cluster analysis is an important problem in data mining and machine learning. In reality, clustering...
In this work-in-progress paper, we present GraphTrees, a novel method that relies on random decision...
We evaluate a similarity measure based on random forests. Existing similarity measure classifies exa...
Measures of similarity play a subtle but important role in a large number of disciplines. For exampl...
Graphs are ubiquitous in many fields of research ranging from sociology to biology. A graph is a ver...
How can we find a good graph clustering of a real-world network, that allows insight into its underl...
Abstract—This work introduces a novel nonparametric density index defined on graphs, the Sum-over-Fo...
This thesis is organized in two parts : the first part focuses on measures of similarity (or proximi...
This work introduces a novel way to identify dense regions in a graph based on a mode-seeking cluste...
Information networks, such as biological or social networks, contain groups of related entities, whi...
Information networks, such as biological or social networks, contain groups of related entities, whi...
Graph clustering methods such as spectral clustering are defined for general weighted graphs. In mac...
In this thesis, we propose many developments in the context of Structural Similarity. We address bot...
Graph clustering methods such as spectral clustering are defined for general weighted graphs. In mac...
We study clustering over multiple graphs- each encoding a distinct set of similarity relationships (...
Cluster analysis is an important problem in data mining and machine learning. In reality, clustering...
In this work-in-progress paper, we present GraphTrees, a novel method that relies on random decision...
We evaluate a similarity measure based on random forests. Existing similarity measure classifies exa...
Measures of similarity play a subtle but important role in a large number of disciplines. For exampl...
Graphs are ubiquitous in many fields of research ranging from sociology to biology. A graph is a ver...
How can we find a good graph clustering of a real-world network, that allows insight into its underl...