Finding Regular Simple Paths in Graph Databases
SIAM Journal on Computing
A partial k-arboretum of graphs with bounded treewidth
Theoretical Computer Science
Making large-scale support vector machine learning practical
Advances in kernel methods
Elements of the Theory of Computation
Elements of the Theory of Computation
Formal-Language-Constrained Path Problems
SIAM Journal on Computing
Efficiently mining frequent trees in a forest
Proceedings of the eighth ACM SIGKDD international conference on Knowledge discovery and data mining
Frequent Sub-Structure-Based Approaches for Classifying Chemical Compounds
ICDM '03 Proceedings of the Third IEEE International Conference on Data Mining
A survey of kernels for structured data
ACM SIGKDD Explorations Newsletter
Kernel Methods for Pattern Analysis
Kernel Methods for Pattern Analysis
The Parameterized Complexity of Counting Problems
SIAM Journal on Computing
Cyclic pattern kernels for predictive graph mining
Proceedings of the tenth ACM SIGKDD international conference on Knowledge discovery and data mining
Frequent subgraph mining in outerplanar graphs
Data Mining and Knowledge Discovery
Network ensemble clustering using latent roles
Advances in Data Analysis and Classification
A new protein graph model for function prediction
Computational Biology and Chemistry
| Hi-index | 0.00 |
The cyclic pattern kernel (CPK) is a powerful graph kernel based on patterns formed by simple cycles of labeled graphs. In a recent work, we proposed a method for computing CPK which is restricted to graphs containing polynomial number of simple cycles. In this work, we present two approaches relaxing this limitation. We first show that for graphs of bounded treewidth, CPK can be computed in time polynomial in the number of cyclic patterns, which in turn can be exponentially smaller than that of simple cycles. We then propose an alternative CPK based on the set of relevant cycles which is known to be enumerable with polynomial delay and its cardinality is typically only cubic in the number of vertices. Empirical results on the NCI-HIV dataset indicate that there is no significant difference in predictive performance between CPK based on simple cycles and that based on relevant cycles.