I'm a research scientist at Samsung - SAIT AI Lab, Montreal, which is located within Mila. My current research focuses on deep learning with structured objects like sets and their equivariance properties.

[scholar] [github] [email] [phd thesis]

Featured

• New ICLR 2022 paper and video on multiset-equivariance!

Papers

• NEW: Unlocking Slot Attention by Changing Optimal Transport Costs.
  Yan Zhang*, David W. Zhang*, Simon Lacoste-Julien, Gertjan J. Burghouts, Cees G. M. Snoek. ICML 2023. Make slot attention more powerful by taking an optimal transport perspective.

• NEW: CrossSplit: Mitigating Label Noise Memorization through Data Splitting.
  Jihye Kim, Aristide Baratin, Yan Zhang, Simon Lacoste-Julien. ICML 2023. Improve robust learning under label noise by reducing memorization of noisy labels using a novel training framework

• NEW: Equivariance with Learned Canonicalization Functions.
  Sékou-Oumar Kaba*, Arnab Kumar Mondal*, Yan Zhang, Yoshua Bengio, Siamak Ravanbakhsh. ICML 2023. Make models equivariant by learning to map data to canonical examples.

• Multiset-equivariant set prediction with approximate implicit differentiation.
  Yan Zhang*, David W. Zhang*, Simon Lacoste-Julien, Gertjan J. Burghouts, Cees G. M. Snoek. ICLR 2022. A better permutation-equivariance property for set prediction.

• Better set representations for relational reasoning.
  Qian Huang*, Horace He*, Abhay Singh, Yan Zhang, Ser-Nam Lim, Austin Benson. NeurIPS 2020. Set-structured latent spaces improve generalisation and robustness.

• Deep set prediction networks.
  Yan Zhang, Jonathon Hare, Adam Prügel-Bennett. NeurIPS 2019. To predict a set from a vector, use gradient descent to find a set the encodes to that vector.

• FSPool: Learning set representations with featurewise sort pooling.
  Yan Zhang, Jonathon Hare, Adam Prügel-Bennett. ICLR 2020. Sort in encoder and undo sorting in decoder to avoid responsibility problem in set auto-encoders.

• Learning Representations of Sets through Optimized Permutations.
  Yan Zhang, Jonathon Hare, Adam Prügel-Bennett. ICLR 2019. Learn how to permute a set, then encode permuted set with RNN to obtain a set representation.

• Learning to count objects in natural images for visual question answering.
  Yan Zhang, Jonathon Hare, Adam Prügel-Bennett. ICLR 2018. Enabling visual question answering models to count by handling overlapping object proposals.