On-line error monitoring for several data structures

Abstract

We present several examples of programs which efficiently monitor the answers from queries performed on data structures to determine if any errors are present. Our paper includes the first efficient on-line error monitor for a data structure designed to perform nearest neighbor queries. Applications of nearest neighbor queries are extensive and include learning, categorization, speech processing, and data compression. Our paper also discusses on-line error monitors for priority queues and splittable priority queues. On-line error monitors immediately detect if an error is present in the answer to a query. An error monitor which is not on-line may delay the time of detection until a later query is being processed which may allow the error to propagate or may cause irreversible state changes. On-line monitors can allow a more rapid and accurate response to an error.<>