Hesam Asadollahzadeh

Hesam Asadollahzadeh

B.Sc. Student

Home

Greetings! I am a last-year B.Sc. student in the ECE department at the University of Tehran. Currently, I am a research scientist at Wellcome Sanger Institute, University of Cambridge, UK, supervised by Dr. Mo Lotfollahi. My interests lie in the field of trustworthy ML, with a particular focus on distributional and adversarial robustness of deep networks.

In addition to my primary research, my interests also include deep generative models, computational biology, and self-supervised learning. Currently, I am working on developing a fair model with disentangled latent space for making causal counterfactual predictions. Additionally, we are continuing our research on group robustness and fairness at MLL & RIML.

News

  • [12/2023] Our pre-print "Annotation-Free Group Robustness via Loss-Based Resampling" is now public at arXiv [Poster].
  • [09/2023] Thrilled to share that the following papers have been accepted to ICCV 2023 - OOD Generalization in Computer Vision Workshop🎉:
    • Data-Driven Annotation-Free Group Robustness Across Extremely Unbalanced Group Sizes
      Ghaznavi M., Asadollahzadeh H., Yaghoubi H., Hosseini F., Rohban M., Soleymani M.
      Paper, Slides
    • Evaluating Robustness of Pre-Trained Deep Neural Networks Against Spurious Correlations
      Taherkhani M., Hoseinpour A., Hosseini F., Asadollahzadeh H., Soleymani M.
      Paper, Slides
  • [05/2023] Joined the Machine Learning Lab (MLL) at Sharif University of Technology, conducting research under the supervision of Dr. Mahdieh Soleymani Baghshah and Dr. Mohammad Hossein Rohban in collaboration with Robust and Interpretable Machine Learning Lab (RIML). The research focuses on exploring spurious correlations, examining the robustness of deep models, and enhancing their interpretability.
  • [11/2022] Joined Dr. Mo Lotfollahi's research team at the Wellcome Sanger Institute, actively contributing to research focused on the application of deep learning in computational biology.
  • [06/2022] Joined the Data Analytics Lab at the University of Tehran, actively engaged in research that emphasizes the application of graph neural networks (GNNs) and using time series analysis for human performance recognition.
  • [07/2021] Ranked in the top 5% of GPA among all computer engineering students (19.38/20.0 or 4.0/4.0).
  • [08/2019] Ranked in the top 0.2% among nearly 60,000 students in the Iranian University Entrance Exam (Konkour).

Selected Talks

Deep Generative Models, Pushing the Limits of Creativity

May 24, 2023

Selected Projects (Full List)

BERT_picture

Transformers, BERT, and BEIT

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,
BEIT: BERT Pre-Training of Image Transformers,
Neural Networks & Deep Learning,
Dr. Ahmad Kalhor,
Autumn 2022

Github, Report (Persian)
BERT_picture

Generative Adversarial Networks (DCGAN, AC-GAN, Wasserstein Loss, and WGAN)

Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks,
Conditional Image Synthesis With Auxiliary Classifier GANs,
Wasserstein GAN (WGANs) (Arjovsky et al. 2017),
Neural Networks & Deep Learning,
Dr. Ahmad Kalhor,
Autumn 2022

Github, Report (Persian)