Abstract. General-purpose exhaustive graph mining algorithms have seldom been used in real life contexts due to the high complexity of the process that is mostly based on costly isomorphism tests and countless expansion possibilities. In this paper, we explain how to exploit gridbased representations of problems to efficiently extract frequent grid subgraphs and create Bag-of-Grids which can be used as new features for classification purposes. We provide an efficient grid mining algorithm called GriMA which is designed to scale to large amount of data. We apply our algorithm on image classification problems where typical Bagof-Visual-Words-based techniques are used. However, those techniques make use of limited spatial information in the image which could be beneficial to obtain more discriminative features. Experiments on different datasets show that our algorithm is efficient and that adding the structure may greatly help the image classification process.

A 2D grid is a particular geometric graph that may be used to represent any 2D regular structure such as, for example, pixel grids, game boards, or cellular automata. Pattern mining techniques may be used to automatically extract interesting substructures from these grids. 2D+t grids are temporal sequences of grids which model the evolution of grids through time. In this paper, we show how to extend a 2D grid mining algorithm to 2D+t grids, thus allowing us to efficiently find frequent patterns in 2D+t grids. We evaluate scale-up properties of this algorithm on 2D+t grids generated by a classical cellular automaton, i.e., the game of life, and we show that the extracted spatio-temporal patterns may be used to analyze this kind of cellular automata.