CCS Seminar
Friday - May 4, 2007
12:00 noon
Physics Research Building - Room 595
Professor Gail Carpenter
Cognitive & Neural Systems - Boston University

"Information Fusion and Hierarchical Knowledge Discovery"

Image fusion has been defined as "the acquisition, processing and synergistic combination of information provided by various sensors or by the same sensor in many measuring contexts." (Simone et al., Information Fusion, 3, 2002, p. 3) When multiple sources provide inconsistent data, such methods are called upon to select the accurate information components. As quoted by the International Society of Information Fusion: "Evaluating the reliability of different information sources is crucial when the received data reveal some inconsistencies and we have to choose among various options."

This talk will address a complementary and novel aspect of the information fusion problem, seeking to derive consistent knowledge from sources that are paradoxically both inconsistent and accurate. This is a problem that the human brain solves well. A young child who hears the family pet variously called Spot, puppy, dog, dalmatian, mammal, and animal is not only not alarmed by these conflicting labels but readily uses them to infer functional relationships that are never explicitly specified.

An ARTMAP neural network derives hierarchical knowledge structures from nominally inconsistent training data. The overall pattern of distributed predictions reveals a hierarchy which guides the production of consistently layered knowledge. Even though no inter-class relationships, links, or probabilities are provided during training, the system derives relationship rules, confidence estimates, equivalence classes, and hierarchical structures.

http://ccs.bu.edu