Where does pattern recognition collide?

The automatic assignment of an input, such as an image, a sound, or a vector of measurements, to one of a set of categories whose boundaries have been learned from labelled data. Bishop's 2006 Pattern Recognition and Machine Learning treats this as a statistical problem: estimate the probability that a feature vector belongs to each class, then pick the most likely one. The referent is a computation an algorithm performs on numerical features and scores by its error rate, not a process in a human mind, a founder's insight about a market, or a clinician's judgment about a patient.[1]

Unit

One classified input, a feature vector assigned to a category

Recovery means

Not applicable

Measured by

Evaluated by classification accuracy, error rate, and confusion matrices on held-out test data

The mental process by which incoming sensory information is matched to representations already held in memory, so that a shape becomes a letter or a face becomes a familiar person. Neisser's 1967 Cognitive Psychology gathered the competing accounts of how this happens, from template matching to feature analysis, and made pattern recognition a central topic of the field. The referent is an inferred operation of the human perceptual system, studied through reaction times and errors, and it differs from a machine's classifier, an entrepreneur's read of a market, or a physician's diagnostic reasoning.[2]

Unit

One act of recognising a stimulus as a familiar pattern

Recovery means

Not applicable

Measured by

Inferred from recognition accuracy, reaction time, and characteristic errors in perception experiments

The cognitive act by which an entrepreneur perceives a connection among seemingly unrelated changes in technology, markets, or society and reads them as a meaningful pattern that signals a new business opportunity. Baron's 2006 Opportunity Recognition as Pattern Recognition proposed that spotting opportunities is a matter of connecting the dots between events others see as separate. The referent is a founder's interpretive leap about where markets are heading, studied in management research, not a perceptual match in the mind, a coded classifier, or a clinical judgment about a patient.[3]

Unit

One recognised opportunity, a link drawn among dispersed events

Recovery means

Not applicable

Measured by

Studied through founder interviews, case histories, and experiments contrasting expert and novice entrepreneurs

The rapid, largely non-analytic process by which an experienced clinician recognises a patient's presentation as an instance of a known disease, matching the signs and symptoms to stored illness scripts built from past cases. Norman and Eva's 2010 Diagnostic error and clinical reasoning described how much of expert diagnosis rests on this fast pattern recognition rather than deliberate rule-following, and how it can both succeed and mislead. The referent is a physician's judgment about a real patient, studied in medical education, and it is not a machine's classifier, a perceptual match of a stimulus, or a founder's read of a market.[4]

Unit

One diagnosis, a patient's presentation recognised as a disease

Recovery means

Not applicable

Measured by

Studied by comparing expert and novice accuracy, diagnostic speed, and error rates on clinical cases