Evolutionary computation, one of the fastest growing areas of computer science, involves simulating the essential processes of Darwinian evolution on a computer in order to optimize solutions to a particular problem. The computer holds a population of candidate solutions and through a process of random variation and selection, iteratively improves the quality of the solutions through competition. The Darwinian process is carried out in fast-time so that many thousands of generations can be evaluated, often in minutes or even seconds on standard PCs. Natural Selection, Inc.^® has unique expertise in devising and designing appropriate representations, variation operators, selection procedures to match the task at hand. This can offer an invaluable advantage. Natural Selection, Inc.'s^® president, Dr. Lawrence J. Fogel, helped pioneer the use of evolutionary computation in the 1960s and the technical staff remains ahead of the curve in advancing the state of the art.

The Valuated State Space ^®approach, invented by Natural Selection Inc.’s president, Dr. Lawrence J. Fogel, is a means for distilling the purpose to be achieved in quantitative terms. Problems often go unsolved because they are left ill-defined, or even undefined. The first step to tackling a difficult problem is to understand the problem. The Valuated State Space^® approach provides a detailed, hierarchic description of the parameters of concern, their relative importance, and the manner in which achievement with respect to one dimension can interact with achievement with respect to another dimension. Most solutions to problems never address the real problem, but rather only some extremely simplified description of it, often in terms of least squared error or linear constraints. The Valuated State Space^® approach, when coupled with evolutionary computation, neural networks, fuzzy systems, and other technologies, provides a means for generating specific solutions for your specific problem.

Neural networks are computer models loosely based on how the brain works by distributing knowledge across a network of neurons that are connected via adjustable weights. Such neural models can be extremely useful for pattern recognition, in that they can be trained to detect complex relationships between inputs even when the statistical distributions of those inputs is unknown. Some applications include detecting coding versus noncoding regions of a genetic sequence, detecting a malignancy in a mammogram, or classifying the type of signal sensed when listening to the ocean.

Fuzzy systems are based on the mathematics of linguistics. They are used to capture ideas such as "close" and "heavy" which are subject to interpretation and may not be immediately suitable for distilling into an algorithm. These logic systems are often applied to control problems where a rule might be "push firmly on the yoke as you get near 1000 feet in altitude." Such a rule is far more useful than something like "push on the yoke with a force of 1.2 Newtons when you reach 991.2 feet in altitude." Fuzzy logic provides a means for handling imprecise statements so that they can be interpreted by both the computer and the human expert.

Connect™ Tools:  Natural Selection, Inc.^® has developed Connect™, a commercially available set of software tools that incorporate the core technologies of our expertise. The software runs on standard desktop PCs and can evolve and optimize solutions in scheduling, routing, pattern recognition, forecasting, clustering, and other areas of application, and incorporates evolutionary computation, neural networks, the Valuated State Space^® approach to quantifying objectives, and more. For information on licensing Connect™, or extending Natural Selection, Inc.'s software tools for your application, contact Dr. David Fogel, CEO.