Software & Algorithms
North Dakota State University scientists have created a unique asynchronous cellular automaton which is believed to have several distinct advantages over currently available field-programmable gate arrays (FPGAs). These cellular automata are easily scaled from small circuits to large computing arrays.
This invention involves a Computer Aided Detection (CAD) model that is designed to diagnose Pulmonary Embolism (PE) from CT image information data sheet. This high performance, classification system, includes a Nearest Neighbor Vote based classification and a Local Decision Boundary based classification combined with an evolutionary algorithm for parameter optimization and a vertical data structure for efficient processing. The invention was developed as a solution for the ACM KDD Cup Competition 2006, and won task 3 of that competition.
Vertical Set Inner Product Technology (VSIP) provides an order of magnitude speed-up to clustering numeric data (both full-space clustering and subspace clustering onto any subspace—including oblique subspaces) by providing a horizontal calculation across vertical predicate-tree datasets. The ability to scale the computations to very large data sets is the significant improvement over traditional methods. Can be used in conjunction with P-Tree (Peano count) models and associated algorithms for datamining applications.
Standard data mining techniques have been sufficient for some areas of information analysis where the datasets are small enough that analysis can be performed relatively quickly and efficiently. However, these standard data mining techniques, such as association rule mining (ARM), have not been as successful in areas such as bioinformatics, nanotechnology, VLSI design, and spatial data, which each have extremely large data sets and where mining implicit relationships among the data can be prohibitively time-consuming. These NDSU-developed data mining technologies are designed specifically for organizing extremely large datasets into an efficiently usable form. The organizational format of the data takes into account the fact that different bits of data can have different degrees of contribution to value. For example, in some applications, high-order bits alone may provide the necessary information for data mining, making the retention of all data unnecessary. The organizational format used also recognizes the need to facilitate the representation of a precision hierarchy. That is, a band may be well represented by a single bit or may require eight bits to be appropriately represented.