Production Matching for Large Learning Systems
One of the central results of AI research in the 1970’s was that to achieve good performance, AI systems must have large amounts of knowledge. Machine learning techniques have been developed to automatically acquire knowledge, often in the form of if-then rules (productions). Unfortunately, this often leads to a utility problem | the \learning” ends up causing an overall slowdown in the system. This is because the more rules a system has, the longer it takes to match them against the current situation in order to determine which ones are applicable.