ABSTRACT
A machine learning system is one that automatically improves with experience, adapts to an external environment, or detects and extrapolates patterns. An appropriate machine learning technology could relieve the current economically dictated one-size-fits-all approach to application design. Help systems might specialize themselves to their users to choose an appropriate level of response-sophistication; portables might automatically prefer the most context-appropriate method to minimize power consumption; compilers may learn to optimize code to best exploit the processor, memory, and network resources of the machine on which they find themselves installed; and multimedia delivery systems might learn reliable time-varying patterns of available network bandwidth to optimize delivery decisions. The potential benefits to computer science of such abilities are immense and the opportunities ubiquitous. Machine learning promises to become the fractional horsepower motor of the information age.
