Giter Site home page Giter Site logo

feyzaakyurek / subspace-reg Goto Github PK

View Code? Open in Web Editor NEW
24.0 24.0 2.0 68.24 MB

Code for the ICLR2022 paper on Subspace Regularization for few-shot class incremental image classification

License: MIT License

Python 83.67% Shell 8.97% Jupyter Notebook 7.37%
class-incremental-learning computer-vision continual-learning few-shot-learning image-classification incremental-learning machine-learning subspace-learning

subspace-reg's People

Contributors

ekinakyurek avatar feyzaakyurek avatar hobbitlong avatar wangyueft avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar

subspace-reg's Issues

A question of the function 'get_projected_weight()' in class LangPuller

Hi, thanks for your inspiring work and detailed doc.

  def get_projected_weight(self, pull, base_weight, weights):
      tr = torch.transpose(base_weight, 0, 1)
      Q, R = torch.qr(tr, some=True) # Q is 640x60
      mut = weights @ Q # mut is 5 x 60
      mutnorm = mut / torch.norm(base_weight, dim=1).unsqueeze(0)
      return mutnorm @ base_weight

I wonder if there is a mistake when getting the projection vector 'm'. In my opinion, 'mut' is the coordinate of the orthogonal basis and the projection should be derived from mut @ Q instead of mut @ base_weight. Please let me know if I were wrong.

Thanks a lot!

Question about QR decomposition

Hi,

I have a question regarding the orthogonal basis that is obtained by QR decomposition. The QR decomposition is defined like this:
image
Why did you not take the whole "Q" as the basis but only use "P_C^(0)"?
image

How to run and reproduce experimental results

Hi,
Thanks for this great study. I would like to reproduce you experimental results but I am not sure which file to run to get the experimental results. I followed the instructions from the README file for downloading and extracting files. I ran the train_supervised.py file but got the following error:
FileNotFoundError: [Errno 2] No such file or directory: './data/miniImageNet\miniImageNet_category_split_train_phase_train.pickle'
It would be amazing if you could share how to use your code for reproducing the experiment results or using it for another datasets.
Thank you so much!

A question of the determine statement 'if [[ $cnt -eq $SLURM_ARRAY_TASK_ID ]]' in sample scripts

Hi, thanks for your inspiring work and detailed doc.
When I ran the sample scripts following your guide, I found that the value of SLURM_ARRAY_TASK_ID is unknown. As a result, no log files were output. So I commented out the whole line if [[ $cnt -eq $SLURM_ARRAY_TASK_ID ]]; then and gave a random value to SLURM_ARRAY_TASK_ID, then everything seemed to be work properly.
I wonder whether the value of SLURM_ARRAY_TASK_ID and the determine statement in sample scripts is necessary or not. Please let me know if I miss anything important.
Thanks a lot!

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.