iKintosh / iKintosh/GaborNet

Parameters of Gabor functions are NOT learnable

Open
#30 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
107
Forks
28
PR merge metrics
No merged PRs in 30d

Description

GaborConv2d implementation does not change the filter parameters (freq, sigma, theta, etc.) during training.

Tested with torch 2.3.0+cu121

If you create a simple network with GaborConv2d first layer, and check the values ​​and gradients for sigma, theta, etc. during training. Values ​​of these parameters will not change, and their gradients will be None.

Contributor guide

No contributing guide indexed for this repository

Research direction

Locate the GaborConv2d implementation and reproduce the issue with a simple network using torch 2.3.0+cu121. Inspect how freq, sigma, theta, and related parameters are defined and used during training. Done means these parameters receive gradients and their values change after optimizer steps.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
computer-vision, machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.