Abstract. We present a method to specify software for a special kind of safetycritical embedded systems, where sensors deliver low-level values that must be abstracted and pre-proc...
Abstract-- Identification of output error models from frequency domain data generally results in a non-convex optimization problem. A well-known method to approach the output error...
Modern machine learning techniques provide robust approaches for data-driven modeling and critical information extraction, while human experts hold the advantage of possessing hig...
Local minima and plateaus pose a serious problem in learning of neural networks. We investigate the hierarchical geometric structure of the parameter space of three-layer perceptr...
— Active contours have been widely used as image segmentation methods. The use of level set theory has provided more flexibility and convenience for the implementation of active...