The following is a brief summary of our e program, for greater detail go to e Algorithm.
SDI's work in evolutionary computing is centered around the development of our e program. e is an advanced Evolutionary Algorithm (EA) Tool designed to solve very difficult problems. e uses evolutionary processes to discover relationships, functions, and programs from the inherent information in a set of "training" data.
Based on user supplied examples of the problem (the training data), e evolves generalizations in the form of a straightforward, procedural program. e goes beyond the simple paradigm of evolutionary algorithms by encapsulating these programs into reusable components and storing them in a library which represents accumulated learning. This library of past learning is available to solve new problems.
You can tackle your difficult problems with the EA that gets smarter as it learns from experience. e is a mature evolutionary algorithm tool that has been used to solve problems in industry and government for nearly ten years. Potential applications are limited only by your imagination. e may be applied to problems in:
Engineering |
Economics |
Signal Processing |
Control |
Pattern Recognition |
Finance |
Distribution |
Sensor Fusion |
Management |
Mathematics |
Science |
Numerical Analysis |
e incorporates numerous EA features which allow users to manipulate the program to their needs.
Major features of our e program:
Evolves programs and coefficients. |
Multiple Ecosystems |
Entities migrate and intermarry among ecosystems along preset paths or make your own arbitrarily complex migration patters. |
Synthetic Annealing (Hill Climbing) integrated into the evolutionary algorithm. |
The solution length and number of coefficients in the solution are determined at run-time. |
Programs are automatically culled for instructions that do not change the outcome. |
You can provide program coefficients or allow e to create its own. |
Populations can easily be saved and reused. |
Libraries of past solutions are automatically applied to the current problem. Current solutions augment the library of stored learning. |
Build multiple libraries for different problem areas. |
Other features that "e" offers:
Configure e to solve floating point problems or make binary decisions. |
Configure e to evolve stock trader programs that output buy, sell, and hold signals |
Use decode utility to directly apply an e program to new input data. |
User can constrain program size. |
User can constrain coefficients in number and range. |
All reproduction parameters are configurable. |
Reproduction operators include crossover, X, blending, mutation, and annealing. |
Configure e to terminate according to any combination of fitness, number of generations, and elapsed time. |
Rich set of configurable program primitives, including arithmetic, logical, conditional, and indirect operators. |
Configurable number of registers available to an e program. |
Scoring options available. |
|
