In many statistical learning problems, it is desired that the optimal solution conform to an a priori known sparsity structure represented by directed acyclic graph. Inducing such structures means of convex regularizers requires nonsmooth penalty functions exploit group overlapping. Our study focuses on evaluating proximal operator latent overlapping lasso developed Jacob et al. in 2009. We imp...

Graph neural networks (GNN) have been proven to be mature enough for handling graph-structured data on node-level graph representation learning tasks. However, the pooling technique expressive graph-level is critical yet still challenging. Existing methods either struggle capture local substructure or fail effectively utilize high-order dependency, thus diminishing expression capability. In thi...

Current Genetic Algorithms can efficiently address order-k separable problems, in which the order of the linkage is restricted to a low value k. Outside this class, there exist hierarchical problems that cannot be addressed by current genetic algorithms, yet can be addressed efficiently in principle by exploiting hierarchy. We delineate the class of hierarchical problems, and describe a framewo...