The Perils and Promise of Evolutionary Computation on Wall Street

Abstract

Real-world examples of evolutionay computation technologies in use in institutional asset management are scarce. The passage of time and the intellectual generosity of the sponsoring firm allow publication of this report. It underscores the problems of applying techniques that rely on massive computation, such as evolutionary computation, neural networks, and knowledge discovery in financial settings. Much of the published work in these areas exhibits at least one of the sins of data mining. Statistical techniques developed in a era of manual or expensive machine computation can be fooled as we push the edges of Moore's Law. Additional measures are needed to find financially and economically meaningful solutions.