Grigoris

Grigoris

Code

As a proponent of open source code, below you can find the source code of several publications I have worked on:

  • On Polynomial Networks:
    • The source code from our work (ICLR'24) on constructing polynomial networks that perform on par with ViT, DeIT (for recognition).
    • The source code from our work (CVPR'23) on regularization of the polynomial networks (for recognition).
    • The source code from our work (NeurIPS'22) on extrapolation and spectral bias of networks with Hadamard product.
    • The source code from our work (NeurIPS'22) on verification of the polynomial expansions (for recognition).
    • Introductory jupyter notebooks from our tutorials at CVPR'22, AAAI'23. I suggest checking out the tutorials if you are new to working with polynomial nets, as they are designed to be educational.
    • The source code from our work (ECCV'22) on polynomial expansions for recognition.
    • The source code from our work (NeurIPS'21) on conditional generation.
    • The source code from our work (CVPR'20/PAMI'21) on Π-nets. Source code for diverse experiments: data generation, data classifcation, face recognition and non-euclidean representation learning.
  • On robustness and verification:
    • Source code for our work (on ICML'24) on a character-level adversarial attack for language models.
    • Source code for our work (on ICLR'24) on NAS for adversarially robust models.
    • Source code for our work (on ICLR'24) on avoiding catastrophic overfitting in single-step adversarial training.
    • Source code for our paper (TMLR'23) on improving the worst class under adversarial training.
    • Source code for our work (ICLR'19/IJCV'20) on robust generative models.
  • Source code for our work (on NeurIPS'24) on scaling the number and specialization of experts in Mixture-of-Expert architectures.
  • Source code for our work (on ICML'24) on studying the interpolation abilities of diffusion models.
  • Source code for our work (on NeurIPS'23) on the maximum independent set (MIS).
  • Source code for our paper (TMLR'23) on federated learning with covariate shift.
  • Source code for our review paper (Proceedings of the IEEE, 2021) on tensor methods.
  • [older] Source code from our work on deformable tracking.
  • [older] Jupyter notebooks from our review paper on deformable tracking.

 

During my PhD, I had also developed and used this library with simple functions on python.