2022
Virtual Tasks but Real Gains: Improving Multi-Task Learning
26th International Conference on Pattern Recognition (ICPR), pp. 4829–4836.
Research
Learning Systems guided by Principled Ideas
My research explores machine learning approaches inspired by how intelligent systems learn continually, use similarity and predict future observations.
Research overview
Principled Learning
My doctoral research investigated four learning principles: continual learning, recency, similarity-based learning and predictive coding. The work studied how these principles can improve learning under delayed feedback, multi-task learning and anomaly detection in multivariate time-series data.
- Continual LearningLearning from streams of data when feedback may be delayed or incomplete.
- Multi-task LearningUsing task similarity and synthetic auxiliary tasks to improve performance.
- Predictive CodingLearning useful representations by predicting future observations.
- Computer VisionApplications in image understanding, medical imaging and adaptive models.
Publications
Selected and peer-reviewed work
Publications are listed in reverse chronological order. External links lead to publisher or open-access records.
2022
Contrastive Predictive Coding for Anomaly Detection in Multi-variate Time Series Data
arXiv preprint arXiv:2202.03639.
2020
2018
Improving Domain-specific SMT for Low-resourced Languages Using Data from Different Domains
Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018).
2016
2016
Research collaboration and student projects
I welcome conversations around Continual Learning, Representation Learning, Computer Vision, Educational Technology and Applied AI systems.