Fiddling with settings to make programmes run faster
Everyday there are complex computer programmes shaping the world around us. They’re used to design chips in the computers and phones. They schedule the trains, buses and flights we take daily. They plan the logistics of getting that Amazon package from a warehouse in the UK to your front door quickly and at a time that suits you.
These programs are called ‘solvers’ and their job is to solve and optimise problems. These solvers can sometimes take hours, days or even weeks to solve these challenging problems. They have a large number of settings which can greatly affect the time it takes to find a solution for these problems.
Tadhg’s research focuses finding the best settings for these solvers. To do this, he races competing settings against each other in parallel and uses a ranking system to rank different settings. This ranking can then be used to choose the best setting to run, and to provide insights into which settings work well and which don’t. His team can also use this information to make new settings by combining good settings to create even better ones. Their results show that this method can greatly reduce the solving time required for these incredibly useful programmes and allow them improve the world around us.
Insight Researchers: Barry O'Sullivan, Tadhg FitzgeraldNon-Insight Researchers: Yuri MalitskyEvent: The International Symposium on Combinatorial Search, Venue: Hotel NH Prague, PragueDate: 16/08/2014Event Type: ConferencePresentation Type: Paper