Nader Asadi

ML Research Engineer, MSc in Machine Learning from Mila.

About

I’m currently a Machine Learning Scientist at Huawei Noah’s Ark Lab, Montreal. I’m working on efficient training/finetuning of large-scale deep learning systems, including the topics of decentralized learning and parameter-efficient tuning. I’m also working on the alignment and reasoning of LLMs, using process-supervised reward models and inference-time intervention.

Previously, I was a graduate student at Mila, working with Eugene Belilovsky and Rahaf Aljundi. I worked on the generalization and continual/decentralized training of deep neural networks.

I was also a research intern at Borealis AI, worked on asynchronous time-series forecasting and regularization techniques for neural temporal point processes.

Research

My research has focused on exploring innovative paradigms for the next generation of large-scale deep learning systems, given any source of supervision. These systems will exhibit continuous growth in modalities (multi-modal learning), tasks (continual learning), and computation (decentralized learning).

Writings

I like sharing my thoughts and writings on machine learning ideas, concepts, and applications on my Blog and Twitter.

News

Apr 24, 2023 PRD got accepted to ICML 2023! :tada:
Feb 6, 2023 I’m serving as a reviewer for ICCV 2023. 🕵️
Jan 9, 2023 I’m joining Borealis AI as a Machine Learning Research Intern. :man_technologist:
Dec 1, 2022 I’m serving as a reviewer for CVPR 2023. 🕵️
May 12, 2022 I’m serving as a reviewer for TPAMI. 🕵️
Mar 2, 2022 Our paper got accepted to CVPR 2022! :tada:

Selected Publications

  1. prd_2023.png
    Prototype-Sample Relation Distillation: Towards Replay-Free Continual Learning
    Nader AsadiMohammadReza Davari, Sudhir Mudur, and 2 more authors
    In ICML, 2023
  2. cvpr_2022.png
    Probing Representation Forgetting in Supervised and Unsupervised Continual Learning
    MohammadReza DavariNader Asadi, Sudhir Mudur, and 2 more authors
    In CVPR, 2022
  3. iclr_2022.png
    New Insights on Reducing Abrupt Representation Change in Online Continual Learning
    Lucas CacciaRahaf AljundiNader Asadi, and 3 more authors
    In ICLR, 2022
  4. neurips_2021.png
    Tackling Online One-Class Incremental Learning by Removing Negative Contrasts
    Nader Asadi, Sudhir Mudur, and Eugene Belilovsky
    In NeurIPS Workshop on Distribution Shifts, 2021