Abstract. While injecting weight noise during training has been proposed for more than a decade to improve the convergence, generalization and fault tolerance of a neural network, ...
One of the main challenges for the development of spatial information theory is the formalization of the concepts of space and spatial relations. Currently, most spatial data struc...
One interesting way of accessing collections of multimedia objects is by methods of visualization and clustering. Growing self-organizing maps provide such a solution, which adapts...
The key limiting factor in graphical model inference and learning is the complexity of the partition function. We thus ask the question: what are the most general conditions under...