{"id":2961,"date":"2020-05-14T11:56:31","date_gmt":"2020-05-14T09:56:31","guid":{"rendered":"https:\/\/aist.fh-hagenberg.at\/?p=2961"},"modified":"2020-05-14T11:56:31","modified_gmt":"2020-05-14T09:56:31","slug":"the-strange-dissapearance-of-oliver","status":"publish","type":"post","link":"https:\/\/aist.fh-hagenberg.at\/index.php\/en\/2020\/05\/14\/the-strange-dissapearance-of-oliver","title":{"rendered":"The strange dissapearance of Oliver"},"content":{"rendered":"

Well it wasn’t really that strange, I was just on Sabbatical for one semester for the purpose of moving forward on my PhD which I am usually doing in my free time. But now that you already fell for the clickbait here is what I did during that time. First of all, I organized everything I wanted to do beforehand. I met Bill Langdon at ICSE<\/a> 2019 and simply asked if I could come visit and collaborate on Genetic Improvement. I also messaged some Professors, most of which replied favorably, leading to smaller exchanges that I won’t go into detail about.Jason Steen<\/a> and I met by sheer coincidence at an HL7 Event and since he needed someone who knows their way around education, and I had a lot of time on my hand when he needed that someone, we planned several education sessions that eventually took place at the HL7 Workgroup Meeting in Sydney.<\/p>\n

So for anyone looking forward to their own sabbatical – Plan everything before<\/em> you go on sabbatical (organizing most of this took months), and also make sure you don’t overload. All of this took a lot of work, but ultimately I learned a lot from the people I visited and have a much stronger foundation moving forward on my own work as well as continuing to collaborate with them.<\/p>\n

Research Stay at the University College London (UCL)<\/h3>\n

The visit to UCL was planned with a specific topic in mind – Utilizing Genetic Improvement (GI)<\/a> to update or introduce code constants in code. This is a niche research area in GI, which usually deals with the improvement of the code itself. The approach<\/a> was<\/a> previously<\/a> shown<\/a> to work really well by evolving lookup tables that were made for one mathematical function, to a lookuptable for another function, using the Covariance Matrix Adaption – Evolution Strategy (CMA-ES<\/a>). CMA-ES is a „relatively“ new algorithm in the evolutionary search domain capable of updating all of its configuration values internally during the run. It has been shown to work on a variety of global and local optimization problems.<\/p>\n

So what we did during the collaboration at the Centre for Research in Evolution, Search and Testing (CREST<\/a>), is adapt the approach to become more generic, accepting any given function that can be approximated using Newton Raphson<\/a>. The results are quite successful, for cube root even outperforming the accuracy of the calculation of the Java and C++ standard implementations. You can take a look at the resulting paper, and presentation at EuroGP 2020 here<\/a>. There are also some publications upcoming related to this research field.<\/p>\n

If you get to visit London the next time, be sure to stop by at UCL and visit Jeremy Bentham<\/a>, the spiritual founder of UCL, and so dedicated to his job that he is still attending board meetings even now, close to 200 years after his death.<\/p>\n