Genetic Algorithms

It was one of those computing forums that attracted a diverse cross section of those interested in emerging software technologies. Its members were drawn from a range of backgrounds- the finance sector, the health sector and the defence sector to name but a few. There was interest around neural networks where the necessary software toolkits were rapidly developing. There was also the concept of genetic algorithms which when first encountered certainly made an impact. In particular the association of the technique with the genetic evolution of species introduced the concept of a software technique that could improve a defined system to higher levels of efficiency and function – effectively by trial and error.

There was the widest of scope for applying the technique of genetic algorithms to real world solutions. The key elements of such a system of optimisation included a typical binary number representation of the ‘system’, functions to modify this number and a ‘fitness’ function that represented how well a specific solution functioned. The binary number as a sequence of 1 and 0 digits could be as long as required – say 50 digits as an example. The fitness function could provide a single output parameter value – say between 0 and 10 – or perhaps several distinct outputs which described distinct characteristics of a particular solution. A range of mechanisms were possible to alter the binary sequence if 0 and 1 digits. Single digits at a specific location could change state or a section could be replaced from a separate independent solution.
I recall that I once described the method as a means to design a hospital based on the organisation and proximity of various departments where the fitness function would try to minimise the level of patient flows between departments in terms of numbers and distances travelled. Such a system could readily to structured using the much looked down upon but entirely practical Microsoft BASIC that ran rather well as I recall under MSDOS .There are obvious elements here of machine learning algorithms. The developments that must have taken place in the scope and power of such software systems since the first emergence of these techniques must, however, be staggering indeed. The role played by such techniques in preserving the security of nations and conversely in seeking to compromise the security of other nations can only be wondered at.

Published
Categorised as Computing

By northernlight1

I have interests is a wide range of topics and have written on these and more formal subjects for quite some time. The written word still retains the power to inform and motivate - hopefully constructively and certainly has to be used responsibly in an age of false information trails.