Nightingale Colloquium presents Cynthia Liem
- Tuesday 14 May 2019
- The seminar is targeted at a broad audience, in particular we invite PhD candidates and supervisors involved in the Data Science Research programme as well as colleagues from LIACS and MI to attend. The seminar is organized by the DSO, MI and LIACS.
Niels Bohrweg 1
2333 CA Leiden
Applying machine learning to human-interpreted data: do we learn what we intend to?
Cynthia obtained her PhD in Computer Science at TU Delft, which she combined successfully with a Master's study in Classical Piano Performance at the Royal Conservatoire of The Hague. Her research interests focus on information filtering challenges in large multimodal archives; in this, she is particularly interested in the algorithmic surfacing of information that users would not discover by themselves, as well as questions of validation and validity in data science.
In this lecture, Cynthia will discuss how her interest in music sparked her research questions on filtering, search and recommendation. In doing this, she will present music as a prime example of multimodal, human-interpreted information, and argue how this poses technological challenges in applied machine learning frameworks, from data representation to the evaluation of success. She will present ongoing work in her lab on the interpretability and trustability of deep-learned music representations, while also discussing how her interests led to several interdisciplinary collaborations with researchers in digital humanities and social sciences.