Evolutionary Algorithms (EA's) are well established and proven. They have an excellent track record in making the otherwise intractable simple. There is sound theoretical grounding to prove their speed and accuracy in finding solutions (although practitioners frequently report that their performance exceeds what theory would have predicted) and there are a wide variety of evolutionary methods – not just GA’s.

Parallel Computing
Part of the power of evolution is due to its parallel nature. Computations running in parallel simply run much faster. High performance parallel computing does not always require high-end hardware and large-scale budgets. General Purpose Graphical Processing Units (GPGPU) now contain 100’s of processors on a PC card. By providing hardware and algorithms, we deliver the means to make 50 to 100 times speed-up possible for a wide variety of problems - not just EA's.

There are two things that make Darwinian different..

Firstly, we are evolutionary and parallel algorithm specialists. It means we can build on particular expertise and experiences to become expert at the peculiar skills needed to build and train fast computations. We also continuously keep abreast of new techniques, available software libraries and productivity tools.

Secondly, Darwinian provides some unique solution development and machine learning tools. For some problems these are many orders of magnitude faster than other methods (say, taking minutes to run instead of years) and may provide a break-through in the scaleability of a large class of computational problems.

Darwinian has a number of sub-specialisms. For example, current effort is going into discovering predictive models with GA’s that can be used in a variety of applications from genetic research to industrial informatics. Another thread is using the same core technology to intelligently recognize captured images. Example applications in industry are:

• Discovery and analysis in biological data
• Searching and cognition in unstructured data
• Industrial and control applications
• Search, monitoring and processing with machine vision
• Intelligent inspection
• Vision for Security

We like nothing more than working directly with researchers, developers and end-users on their difficult problems and will provide products and services including: viability assessments, technical consulting, solution design and full lifecycle development and implementation of bespoke solutions.

Please Contact us for more detailed information.

Definition:
Evolutionary computation uses software that converts an otherwise difficult problem by encoding very many plausible solutions into a large population which then has evolutionary pressures applied to it over many generations to breed better versions and ultimately arrive at an answer of value (the answer might be a model to predict behavior, better tuned parameters for a system, a computerized simulation of something in the real world or some likely causal connections that were not known before).
[Back to top]

[ Top of page ]   [ Contact ]  [ Privacy Policy ]