Shell Xu Hu
Coriolis B413École des ponts ParisTech
6-8, Av Blaise Pascal
Champs-sur-Marne 77455
About Me
I am a research scientist at Samsung AI Center Cambridge. Previously, I was a postdoc under the NINAI project, where I was also appointed CTO of Upload AI. I did my PhD at the IMAGINE lab, École des Ponts ParisTech, supervised by Guillaume Obozinski and Nikos Komodakis. I spent good time as a visiting postdoc at Johan Suykens' group at KU Leuven and as an applied scientist intern at Amazon Cambridge with Andreas Damianou and so on.
News
- I have successfully defended my PhD thesis. A copy of the manuscript can be found here.
- Our paper Empirical Bayes Transductive Meta-Learning with Synthetic Gradients is accepted by ICLR 2020. Our code will be released in Xfer.
- Our paper Empirical Bayes Meta-Learning with Synthetic Gradients is accepted by MetaLearn 2019.
- I am a visiting scholar at KU Leuven since July 8th.
- Attended ICML 2019 and CVPR 2019.
- Our paper Variational Information Distillation for Knowledge Transfer is accepted by CVPR 2019.
- Gave a talk at STADIUS, KU Leuven, On Variational Characterization of Mutual Information for Regularizing Deep Learning, Nov 30th 2018.
- Our paper beta-BNN: A Rate-Distortion Perspective on Bayesian Neural Networks is accepted by NeurIPS-BDL 2018.
- Our paper Variational Mutual Information Distillation for Transfer Learning is accepted by NeurIPS-CL 2018.
- I am doing an internship at Amazon Cambridge supervised by Andreas Damianou and Pablo García Moreno.
- Our paper SDCA-Powered Inexact Dual Augmented Lagrangian Method for Fast CRF Learning accepted at AISTATS 2018.
- Gave a poster presentation at KAUST workshop on optimization and big data on fast CRF learning.
- Attended Data Science Summer School 2017 at Ecole polytechnique.
- I was at NIPS 2016 Barcelona.
- I am a volunteer of ICML 2015 at Lille.
- I am the mentor of a GSoC 2014 project for Shogun Machine Learning Toolbox: Structured Output Learning with Approximate Inference. See our IPython notebook by Jiaolong Xu.
- My GSoC 2013 project of general structured output models for Shogun Machine Learning Toolbox is finished. A IPython notebook is available.