Skip to content

Implementation of "Deep Learning in Random Neural Fields: Numerical Experiments via Neural Tangent Kernel"

License

Notifications You must be signed in to change notification settings

kwignb/RandomNeuralField

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Random Neural Fields

Library

  • PyTorch 1.10.2
  • torchvision 0.11.3
  • others can be installed with the following:
pip install -r requirements.txt

Config

  • src/config/config.yaml
DATA:
  NAME: mnist           # Now: mnist or fashion
  CLASS: 10             # number of class of mnist
  DATA_NUM: 100
  SPLIT_RATIO: 0.2

MODEL:
  INPUT_FEATURES: 784   # mnist: 28x28=784
  MID_FEATURES: 1024
  B_SIGMA: 0.1

INITIALIZER:
  TYPE: gaussian        # vanilla, gaussian, withmp, mexican, matern
  R_SIGMA: 0.1          # for receptive field
  S_SIGMA: 0.1          # for correlation
  M_SIGMA: 0.1          # for mexican hat
  NU: 0.5               # for matern kernel

GENERAL:
  EPOCH: 1000           
  GPUS: [1]
  NOTEBOOK: true        # if you use script, switch false

Citation

@article{watanabe2023deep,
  title={Deep learning in random neural fields: Numerical experiments via neural tangent kernel},
  author={Watanabe, Kaito and Sakamoto, Kotaro and Karakida, Ryo and Sonoda, Sho and Amari, Shun-ichi},
  journal={Neural Networks},
  volume={160},
  pages={148--163},
  year={2023},
  publisher={Elsevier}
}

About

Implementation of "Deep Learning in Random Neural Fields: Numerical Experiments via Neural Tangent Kernel"

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published