Generalized Median Graph via Iterative Alternate Minimizations
This addresses a core computational challenge in graph-based machine learning, but it is incremental as it builds on existing methods for median graph computation.
The paper tackles the NP-hard problem of computing a generalized median graph for clustering or classification by proposing an efficient block coordinate descent approach that handles node and edge labeling, showing efficiency in experiments on various datasets.
Computing a graph prototype may constitute a core element for clustering or classification tasks. However, its computation is an NP-Hard problem, even for simple classes of graphs. In this paper, we propose an efficient approach based on block coordinate descent to compute a generalized median graph from a set of graphs. This approach relies on a clear definition of the optimization process and handles labeling on both edges and nodes. This iterative process optimizes the edit operations to perform on a graph alternatively on nodes and edges. Several experiments on different datasets show the efficiency of our approach.