Linh Tran (she/her), Ph.D.

Email | CV | Google Scholar | Twitter

I am a Research Scientist at Meta working on deep generative models, in particular Text-to-Image Diffusion Models. Before joining Meta, I was a Senior Research Scientist at Autodesk AI lab and interned at Google Brain and Samsung AI. Before my time in industry, I spent several years at Imperial College London completing my PhD under the supervision of Maja Pantic.

I am generally interested in deep probablisitc modeling, learning with limited supervision, and representation learning. I am excited to apply these methods to 2D and 3D computer vision and robotics (vision-based, Sim-to-Real transfer). Further, I am also interested in developing interpretable and trustworthy machine learning models. My PhD research focused on learning interpretable and human-controllable representations using deep generative modeling.

I have been volunteering for several non-profit organizations. I was fortunate to be one of the organizers of the Women in Machine Learning (WiML) workshop (co-located at NeurIPS) in 2021. I served as a Board Member for the non-profit organization Women in Machine Learning (WiML) from April 2022 until July 2023. In collaboration with Autodesk Foundation and Bridges to Prosperity,, I led a multi-disciplinary team over six months, analyzing geospatial data and developing deep learning models for remote bridge site identification in 2022.

Outside of work, I am mum to a little toddler called Milo. I enjoy being creative, organising events for friends & family, travelling the world, gardening and yoga.

I am always happy to chat about research and possibilities of collaborations. Feel free to drop me an email.



Selected publications

Cauchy-Schwarz Regularized Autoencoder
L. Tran, M. Pantic and M.P. Deisenroth
Journal of Machine Learning Research, 2022
[ arxiv | jmlr proceedings | code ]
Hydra: Preserving Ensemble Diversity for Model Distillation
L. Tran, B.S. Veeling, K. Roth, J. Swiatkowski, J.V. Dillon, J. Snoek, S. Mandt, T. Salimans, S. Nowozin and R. Jenatton
Uncertainty & Robustness in Deep Learning (UDL), ICML 2020
[ arxiv (long version) | workshop page | workshop paper ]
Disentangling Geometry and Appearance with Regularised Geometry-Aware Generative Adversarial Networks
L. Tran, J. Kossaifi, Y. Panagakis and M. Pantic
International Journal of Computer Vision, pp.1-21, 2019.
[ ijcv open access ]
GAGAN: Geometry-Aware Generative Adversarial networks
J. Kossaifi, L. Tran, Y. Panagakis and M. Pantic
Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018.
[ arxiv | cvpr proceedings ]
Deepcoder: Semi-parametric Variational Autoencoders for Automatic Facial Action Coding.
D. L. Tran*, R. Walecki*, Ognjen (Oggi) Rudovic*, S. Eleftheriadis, B. Schüller and M. Pantic (* denotes equal contribution)
Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017
[ arxiv | iccv proceedings | poster ]

Conference articles

Generalizable Pose Estimation Using Implicit Scene Representations
V. Saxena, K. H. Maleksha, L. Tran* and Yotto Koga* (* denotes equal supervision)
International Conference on Robotics and Automation, 2023
[ project | arxiv | code ]
Masktune: Mitigating spurious correlations by forcing to explore
S. Asgari, A. Khani, F. Khani, A. Gholami, L. Tran, Ali Mahdavi Amiri and Ghassan Hamarneh
Advances in Neural Information Processing Systems (NeurIPS), 2023
[ NeurIPS openreview | arxiv | code ]
SimCURL: Simple Contrastive User Representation Learning from Command Sequences
H. Chu, A. H. Khasahmadi, K. D. D. Willis, F. Anderson, Y. Mao, L. Tran, J. Matejka, J. Vermeulen
International Conference on Machine Learning and Applications (ICMLA), 2023
[ ICMLA proceedings | arxiv ]
Evolving Through the Looking Glass: Learning Improved Search Spaces with Variational Autoencoders
P. J. Bentley, S. L. Lim, A. Gaier and L. Tran
17th International Conference on Parallel Problem Solving from Nature (PPSN XVII), 2022
[ PPSN proceedings | code ]
JoinABLe: Learning Bottom-up Assembly of Parametric CAD Joints
K. D. D. Willis, P. K. Jayaraman, H. Chu, Y. Tian, Y. Li, D. Grandi, A. Sanghi, L. Tran, J. G. Lambourne, A. Solar-Lezama and W. Matusik
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022.
[ arxiv | cvpr proceedings | poster | code | dataset ]
How Good is the Bayes Posterior in Deep Neural Networks Really?
F. Wenzel*, K. Roth*, B.S. Veeling*, J. Swiatkowski, Tran, L., S. Mandt, J. Snoek, T. Salimans, R. Jenatton and S. Nowozin (* denotes equal contribution)
37th International Conference on Machine Learning (ICML), 2020
[ arxiv | icml proceedings | github ]
The k-tied Normal Distribution: A Compact Parameterization of Gaussian Mean Field Posteriors in Bayesian Neural Networks
J. Swiatkowski, K. Roth, B.S. Veeling, L. Tran, J.V. Dillon, S. Mandt, J. Snoek, T. Salimans, R. Jenatton and S. Nowozin
37th International Conference on Machine Learning (ICML), 2020
[ arxiv | icml proceedings ]

Journal articles

ProfPPIdb: pairs of physical protein-protein interactions predicted for entire proteomes
L. Tran, T. Hamp and B. Rost
Plos ONE, 2018
[ bioarxiv | plos one open access | project page ]

Peer-reviewed workshop articles

COIL: Constrained Optimization in Learned Latent Space
P. J. Bentley, S. L. Lim, A. Gaier and L. Tran
Workshop on Enhancing Generative Machine Learning with Evolutionary Computation (EGML), GECCO 2022
[ arxiv (long version) | workshop page | workshop paper | code ]
Weakly-Supervised Group Disentanglement using Total Correlation
L. Tran, S.A. Taghanaki, A.H. Khasahmadi and A. Sanghi
Workshop on Weakly Supervised Learning (WeaSuL), ICLR 2021
[ arxiv (long version) | workshop page | workshop paper ]


Representation Learning for Sequential Volumetric Design Tasks
M. F. Alam, Y. Wang, L. Tran, C.-Y. Cheng and J. Luo
arXiv preprint arXiv:2309.02583 2022
[ arxiv ]
Counterbalancing Teacher: Regularizing Batch Normalized Models for Robustness
S. Asgari, A. Gholami, F. Khani, K. Choi, L. Tran, R. Zhang and A. Khani
arXiv preprint arXiv:2207.01548 2022
[ arxiv ]


Bridges To Prosperity: Geospatial Deep Learning for Remote Site Identification
L. Tran, M. Nadaf, R. Bushinsky, P. S. Alluru, D. Goswami, A. Noriega and J. Shirley
[ pdf | code ]
Interdisciplinary, machine learning pro bono project for remote bridge site identification with Autodesk Foundation and Bridges to Prosperity.
Human-Controllable and Structured Deep Generative Models
L. Tran
Ph.D. Thesis
[ pdf ]