I’m a Ph.D. student at University of Montreal, where I investigate the integration of deep learning models with reinforcement learning methods – research which now falls under the umbrella term « Deep Reinforcement Learning ». In particular, my current research at MILA focuses on dialogue and question answering systems. I’m jointly supervised by Aaron Courville and Yoshua Bengio, and I collaborate extensively with Joelle Pineau at McGill University.
Before starting my PhD, I obtained my master degree MSc. Machine Learning in 2014 from University College of London. Prior to that, I studied theoretical (pure) mathematics and statistics at the University of Copenhagen.
See my personal website for more information: http://www.iulianserban.com
- Generating Factoid Questions With Recurrent Neural Networks: The 30M Factoid Question-Answer Corpus. Chia-Wei Liu, Ryan Lowe, Iulian V. Serban, Michael Noseworthy, Laurent Charlin, Joelle Pineau. Under review for ACL 2016.
- How NOT To Evaluate Your Dialogue System: An Empirical Study of Unsupervised Evaluation Metrics for Dialogue Response Generation. Iulian Vlad Serban, Alberto García-Durán, Caglar Gulcehre, Sungjin Ahn, Sarath Chandar, Aaron Courville, Yoshua Bengio. Under review for ACL 2016.
- Building End-To-End Dialogue Systems Using Generative Hierarchical Neural Network Models. Iulian V. Serban, Alessandro Sordoni, Yoshua Bengio, Aaron Courville, Joelle Pineau. AAAI 2016.
- The Ubuntu Dialogue Corpus: A Large Dataset for Research in Unstructured Multi-Turn Dialogue Systems. Ryan Lowe, Nissan Pow, Iulian Serban, Joelle Pineau. SIGDIAL 2015.
See arXiv for a more complete list of publications: http://arxiv.org/find/cs/1/au:+Serban_I/0/1/0/all/0/1
See GitHub for open source code projects: https://github.com/julianser
I’m happy to co-supervise and collaborate with undergraduate students. Motivated students with a strong record in computer science, mathematics or statistics are very welcome to contact me.
iulian [DOT] vlad [DOT] serban [AT] umontreal [DOT] ca