The following is the article regarding Bees Algorithm.
Software algorithms that perform such analysis and optimisations have existed for a while. But now, researchers Sameh Otri from Syria and Afshin Ghanbarzadeh from Iran, working for their PhDs at Cardiff University’s Manufacturing Engineering Centre, have written an algorithm they claim can outperform those of its competitors in many ways.
And they developed it by modelling the food-foraging behaviour of swarms of honey bees.
There are many types of so-called swarm-based optimisation algorithms (SOAs). Many of them have already taken their cues from nature. Ant Colony Optimisation (ACO) algorithms, for example, emulate the behaviour of ants, which find the shortest path from their nest to a food source by depositing chemicals to guide their search. Genetic Algorithms (GA), on the other hand, are based on natural selection and genetic recombination. Particle Swarm Optimisation (PSO) algorithms are based on the social behaviour of groups of organisms such as the flocking of birds and the schooling of fish.
The new Bees Algorithm developed at Cardiff University was the brainchild of the university’s Professor Duc Truong Pham, OBE, and was unveiled by his PhD student Afshin Ghanbarzadeh and his colleagues at the recent internet-based Innovative Production and Machines Conference, which the MEC hosted as part of its work with the EUfunded Network of Excellence.
For its part in the affair, the team walked off with the accolade of Best Young Researchers.
‘In developing the new algorithm, we tried to model and mimic the behaviour of bees, since the modelling of ant behaviour had been done by previous researchers,’ said Ghanbarzadeh. ‘We closely looked at the foodforaging behaviour and we built an artificial optimisation based loosely around the procedure.’
So how does the Bees Algorithm actually work? ‘When a bee finds a source of nectar, it returns to the hive and performs a dance called the “Waggle Dance” to show other bees the direction, distance and quality of the
flower patch it has come from,’ said Ghanbarzadeh. ‘After the waggle dance has been performed, the bees know how to get back to the most promising flower patches. In this way, the bees can gather food in the fastest, most efficient way possible.’
The MEC team’s Bees Algorithm mimics this behaviour. Initially working with a random number of possible
solutions to a specific problem, it finds several close matches to an optimum solution. Once that has been done, the algorithm dedicates more resources to those close matches to optimise the solution further until an absolute optimum solution as been found.
More specifically, the algorithm combines a neighbourhood search with a random search — as such it can be used for both combinatorial and functional optimisation. Here’s how. First off, it loosely uses the idea that bees may initially leave a hive with no idea where to find the best patch for food.
Like them, the algorithm first looks randomly for solutions across a search space. It then returns the values of the fitnesses of sites it has found in that search space. Sites of the highest fitnesses are then chosen for a neighbourhood search, and the algorithm conducts searches in the neighbourhood of those selected sites, assigning more resources to the task, much in the way that the bee hive allocates more bees to visit the most promising flower patches.
Nevertheless, other possible solutions are not ignored. Just as scout bees are endlessly looking for new sources of food, the algorithm itself also conducts random searches for solutions with high fitnesses while continuously allocating more and more resources to the most promising of the solutions that it initially discovers.
When Ghanbarzadeh and Otri let their new algorithm loose on the spring optimisation problem to test out its mettle, they found that it behaved faster than the GA or ACO algorithms, and required less computational power to boot.
‘The Bees Algorithm found the optimum result faster and more reliably,’ said Ghanbarzadeh. ‘Now, we’d like to test it on more problems to determine the range of problems that it is specifically useful at optimising.’ Otri added: ‘There are many algorithms and some are good at solving a certain set of problems better than other ones, which is why we plan to spend some more time quantifying the behaviour of the Bees Algorithm that we have developed.’
With his postdoctoral work coming to a close, Ghanbarzadeh now plans to return to Chameron University in
southern Iran, where he worked as a lecturer for five years before winning a scholarship to come to Cardiff to study for his PhD.
For his part, Otri also plans to return to his home town of Damascus in Syria, which he left in 2001 to study for his MSc and PhD at Cardiff after winning a scholarship from Damascus University.
Despite the fact that they might soon be thousands of miles away, the two researchers plan to keep in regular touch with the Manufacturing Engineering Centre at Cardiff University.
They want to keep tabs on developments in intelligent optimisations — and spread the word overseas — as they teach the subject to the next generation of students in their own universities.
Billed as the greenest conference on earth, the Innovative Production and Machines Conference brought together delegates from around the world without the need for travel. You can find out more about their work and the conference on the internet by logging on at: http://conference.iproms.org.Source of Article: technologyhorizons.co.uk
0 comments:
Post a Comment