ANOMALY DETECTION MAY BE REGARDED AS A PATTERN RECOGNITION PROBLEM IN WHICH THE EXPECTED IS KNOWN AND THE UNEXPECTED MUST BE DETECTED. THE PRECISE LINE BETWEEN EXPECTED AND UNEXPECTED OFTEN CANNOT BE IDENTIFIED IN ADVANCE BY A SPECIFIC THRESHOLD OR PROBABILITY. AN EXAMPLE FROM THE PHARMACEUTICAL INDUSTRY IS USED TO SHOW HOW MINIMUM DESCRIPTION ANALYSIS CAN BE APPLIED TO THE RECOGNITION OF ANOMALOUS COLOR DISTRIBUTIONS.
