Abstract: The present invention comprises a computer-implemented method of searching large metric space databases. It allows fast near neighbor searches in databases where the data elements in the database are high dimensional and each data element represents a point in a large metric space. Given a query item, which also represents a point in the large metric space, one or more data items in the database which are approximately nearest neighbors of the query item are found. A set of data items is first preprocessed by computing distances between pairs of items and storing links between pairs which are near one another. A search of the database proceeds by following links from item to item, usually by following links to items which are nearest the query Q. In one embodiment, the search terminates upon reaching an item R which is closer to Q than are all the items to which R links.

Abstract: A pattern recognition device having modifiable feature detectors (28) which respond to a transduced input signal (26) and communicate a feature activity signal (30) to allow classification and an appropriate output action (70). A memory (40) stores a set of comparison patterns, and is used by an assigner (66) to find likely features, or parts, in the current input signal (26). Each part is assigned to a feature detector (28[m]) judged to be responsible for it. An updater (42) modifies each responsible feature detector (28[m]) so as to make its preferred feature more similar to its assigned part. The modification embodies a strong constraint on the feature learning process, in particular an assumption that the ideal features for describing the pattern domain occur independently. This constraint allows improved learning speed and potentially improved scaling properties.