Evolving our understanding – genetic algorithms explained

Maptek Evolution takes advantage of evolutionary algorithms to create dynamic, agile mine schedules that maximise value. But what are evolutionary algorithms? In nature, any living thing must fight for survival. If one individual in a population has characteristics that make it more likely to survive, it is more likely to pass on its genes to the next generation.

Over time, a population of individuals will become better suited for their environment. This is the process of natural selection.

A genetic algorithm is an optimisation technique that mimics natural selection to solve complex optimisation problems. The first step of using a genetic algorithm to solve any problem is finding a way to compress a solution to the problem into an ‘encoding’.

Using this definition, a population is a grouping of encodings. Once the problem can be represented in a population, various natural genetic processes can be simulated. This includes the combination of DNA as two parents breed (crossover), as well as the random changes that happen to a single encoding during the breeding process (mutation).

As in nature, genetic processes take time, trial and error… and genetic algorithms do not take shortcuts or make magic intuitive leaps—they just do it all faster. Access to the cloud has provided computing power that makes it seem easy!

Dead ends are quickly eliminated and the seemingly huge problem is relentlessly narrowed down into optimal solutions.

Now that the problem has been defined and modelled in code, your individual solutions fight it out in the cloud learning and competing with each other in a simultaneous processing environment.

Maptek Evolution is aptly named. It uses evolutionary algorithms to produce dynamic, agile mine schedules that allow operations to maximise value without dumbing-down the data. Thousands of scenarios can be rapidly assessed through high performance cloud computing to generate new, better solutions.

At Maptek, the Evolution product development team has been consolidating genetic algorithm code so we can quickly and easily apply it to new projects and share with other development teams. In this way we consolidate the optimisation building blocks to create a standalone genetic algorithm in the same manner as a linear programming library.

Look out for our next blog, which goes a bit deeper into how genetic algorithms work.

Click here to read part 2 of this blog series >