SAVANNAH, Ga.--(BUSINESS WIRE)--Health Discovery Corporation (OTCBB: HDVY) today announced that the U.S. Patent and Trademark Office has issued a notice of allowance of HDC’s patent application covering methods for selection of kernels, i.e., functions or algorithms that are used to transform input data into a different configuration, known as “feature space”, which allows for easier recognition of patterns within the data. The kernels covered by the application’s claims are useful for analysis of data that may possess characteristics such as structure, for example, DNA or protein sequences, spectrographic data, images, documents, graphs, ECG signals, and many others. These kernels are particularly useful where the input data may possess invariances or noise components that can interfere with the ability to accurately extract the desired information. A location-dependent kernel looks for similarity between structures that is relevant to particular locations in the structures, such as a string of text in two different documents. Using the locational kernel, a pair of structures can be viewed according to the similarities of their components. The locational kernel method is not limited to support vector machines (SVMs), but may be used in other kernel-based learning machines. This application also discloses novel feature selection techniques that consider the possibility of structure within and around the “noise vector” in a problem. The inclusion of these features allows the SVM or other kernel-based classifiers to gain the desirable property of noise invariance.
