I am currently recruiting motivated PhD and Msc students. Please send me an informal inquiry to (firstname.lastname@example.org) if you are interested with your CV and optionally a short (1 page max) description of motivations and research interests. In your email subject please put [Recruiting]. Note I am not be able to respond to all emails. Students should demonstrate strong programming and math skills. Previous research experience, particularly in machine learning is a plus. Below I provide an idea and some references for general areas I am currently looking for students in, however I am open to supervising student proposals in other projects in Machine Learning and Computer Vision.
Continual learning, online deep learning, and few shot learning
An exciting open question in deep learning is how to effectively learn in online and non-stationary settings. Continual learning aims to learn new tasks (with potentially few samples) while retaining knowledge of old tasks. A few related publications:
-Aljundi, Caccia, Belilovsky, Caccia et al Online Continual Learning with Maximally Interferred Retrieval
-Caccia, Belilovsky et al Online Learned Continual Compression with Adaptive Quantization Modules
-Hu et al Drinking from a Firehose: Continual Learning with Web-scale Natural Language
-Lin et al Conditional Computation for Continual Learning
Vision and Language
I am interested in computer vision problems at the intersection of natural language processing (for example VQA, Embodied QA)
Alternative Training of Deep Networks
I am interested in developing novel and scalable methods for training deep networks that have improved speed, convergence, simpler parallelization or hardware implementation, online learning, and/or forgetting. Some related publications to give a flavor.
-Belilovsky, Eugene, Michael Eickenberg, and Edouard Oyallon. “Decoupled greedy learning of cnns.” ICML 2020.
-Xu, An, Zhouyuan Huo, and Heng Huang. “On the Acceleration of Deep Learning Model Parallelism With Staleness.” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020.
-Choromanska, Anna, et al. “Beyond backprop: Online alternating minimization with auxiliary variables.” International Conference on Machine Learning. 2019.