I am currently the manager of the Regulated Software Research Centre (RSRC: https://www.dkit.ie/regulated-software-research-centre) at Dundalk Institute of Technology. Prior to this I was a senior researcher in the Natural Computing Research and Applications Group (NCRA: http://ncra.ucd.ie/index.shtml) in UCD. Both outside and inside work, my passion has always fallen somewhere between music and science. My main research interests currently lie in bias and ethics in AI, and examining computational musical creativity, predominantly using evolutionary computational methods.
Qualifications & Awards
PhD in Computer Science
University of Limerick 2005 – 2009 (Graduated Jan 2010)
· 4 year interdisciplinary PhD role encompassing signal processing, music and machine learning
· Coordinated and managed a series of listening tests with the public (>30 participants)
· Published and presented 5 peer-reviewed conference papers
· Lectured, tutored and supervised students at both undergraduate and post-graduate level
Masters in Music and Media Technology
Trinity College Dublin 2002 – 2004
· 2 year taught masters
· Class representative for both years
· 6 month research project including dissertation and composition for live musicians
Evolutionary Computation (CEC), 2015 IEEE Congress on
2015 | Conference paper
Source: Roisin Loughran
Research & Interests
I am keenly interested in identifying and mitigating against unwanted biases from AI. I am part of the working group for the IEEE P7003 standard on 'Algorithmic Bias Considerations'.
My other interests lie in examining computational musical creativity, predominantly with evolutionary computational methods. Using Grammatical Evolution I have created a number of variations of an evolutionary music generative systems which I have (at least personally) termed the ‘Composing Pony’ with varying fitness functions. This is written in Python using the PonyGE implementation of GE and standard python midi libraries. These systems create short MIDI melodies with no a priori key or time signature and with minimal input from the user. Details of these systems can be found in the relevant papers, most of which are available though my ResearchGate page.
In addition to creating generative systems, I’m particularly interested in means of evaluating autonomous creative systems. This is key to determining whether or not a generated artifact was produced by human or AI means.