NEURAL NETWORKS FOR ADAPTIVE THREAT RESPONSE

NEURAL NETWORKS ARE A PROMISING SOLUTION TO SEVERAL PROBLEMS IN THE APPLICATION OF PARALLEL PROCESSING TECHNOLOGY FOR ADAPTIVE RECOGNITIONAND RESPONSE TO COMPLEX PATTERNS. UNFORTUNATELY, IT IS NOT KNOWN HOW TO GET A NEURAL NETWORK TO RECOGNIZE PATTERNS INDEPENDENTLY OF THE VALUES OF THEIR FEATURE PARAMETERS WITHOUT INCORPORATING SPECIAL INVARIANT TRANSFORMS FOR EACH SUCH FEATURE. THIS PROPOSAL INTRODUCES THE CONCEPT OF THE "PARAMETRIC AVALANCHE", WHICH DESCRIBES HOW A NEURAL NETWORK CAN ASSOCIATE PATTERNS THROUGH THE ACTIVATION OF FEATURE PATHWAYS OTHER THAN SIMPLE TIME PROXIMITY. THROUGH THIS MECHANISM, WE DESCRIBE NOT ONLY HOW TO MEASURE THE TRAJECTORY OF AN OBJECT, BUT ALSO HOW TO EXTRAPOLATE THE TRAJECTORY AND GUIDE A COUNTERMEASURE TO INTERCEPT THE TARGET.