I was born and raised in Band, a beautiful village near Urmia City.

email: arman.afrasiyabi@gmail.com

mila-page: https://mila.quebec/en/person/arman-afrasiyabi/

Arman Afrasiyabi

Mila-Québec AI Institute & Université Laval

I am a final year Ph.D. candidate at Mila and Université Laval supervised by Prof. Jean-François Lalonde and Prof. Christian Gagné. I am working on meta-learning, representation learning, and self-supervised learning in low supervision conditions. With my supervisors, we first proposed the idea of associative alignment for few-shot image classification at the European Conference on Computer Vision (ECCV) 2020 to increase the model capacity of the conventional few-shot learning methods. Second, we proposed a mixture model-based representation learning approach at International Conference on Computer Vision (ICCV) 2021 to learn a robust multi-modal feature space in a semi-supervised way which improves few-shot image classification. While working with my supervisors, I finished my third Ph.D. project in collaboration with Prof. Hugo Larochelle. The work proposes a novel representation learning method based on a hybrid CNN/transformer-type architecture, which is under review now.

Before the Ph.D., I was an M.S. student at biomedical and computer engineering at METU under the supervision of Prof. Fatos T. Y. Vural. I was a research assistant at METU-fMRI and worked on several fully-funded projects on medical image analysis using machine learning.

I was also a teaching assistant for two related graduate-level courses: a) Pattern Recognition and Machine Learning (Bioshp, 2008 based), and b) Deep Learning (Goodfellow et al. 2016 based). I am toward publicly sharing my course slides for Bishop's book.



  1. ICCV
    Mixture-based feature space learning for few-shot image classification
    International Conference on Computer Vision, 2021
    1. ECCV Spotlight
      Associative alignment for few-shot image classification
      European Conference on Computer Vision, 2020
      1. ICASSP
        Non-euclidean vector product for neural networks
        Arman Afrasiyabi, Diaa Badawi, Baris Nasir, Ozan Yildiz, Fatos T. Yarman Vural, A Enis Çetin
        International Conference on Acoustics, Speech and Signal Processing, 2018
        1. ICCI*CC
          A sparse temporal mesh model for brain decoding
          15th International Conference on Cognitive Informatics & Cognitive, 2016


  • ICLR-2022, ICCV-2021, 3DV-2021, and MAIS-2021.

presentations and phd exams

  • At ICCV 2021, I am presenting Mixture-based feature space learning for few-shot. [video]
  • In 26, Feb, 2021, I present Transfomers to Prof. Jean-François Lalonde's research group. [my slides]
  • In Feb 2021, I gave a presentation on PMM and Associative Alignment at IID, Université Laval. [my slides]
  • At ECCV 2020, I presented our work “Associative Alignment for Few-shot Image Classification”. [video]
  • In July 2020, I gave a talk on ”Advances in few-shot learning” at Université Laval, IFT 6501. [my slides]
  • In March 2020, I gave a talk on my last reasearch project at IID, Université Laval. [my slides]
  • In Jan 2019, I successfully passed my PhD exams: [presentation] and [proposal].
  • In January 2018, I gave a presentation on Recurrent Neural Network (RNN) [my slides-RNN]
  • , and Neural Turing Machines: NTM: [my slides-NTM].