https://github.com/benkeene/ntk_utils

Tools for the analysis of ANNs.

https://github.com/benkeene/ntk_utils

Science Score: 10.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
  • codemeta.json file
  • .zenodo.json file
  • DOI references
  • Academic publication links
    Links to: arxiv.org
  • Academic email domains
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (4.1%) to scientific vocabulary
Last synced: 10 months ago · JSON representation

Repository

Tools for the analysis of ANNs.

Basic Info
  • Host: GitHub
  • Owner: benkeene
  • Language: Python
  • Default Branch: master
  • Size: 23.5 MB
Statistics
  • Stars: 0
  • Watchers: 0
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Fork of ajacot/NTK_utils
Created about 3 years ago · Last pushed about 3 years ago

https://github.com/benkeene/NTK_utils/blob/master/

# NTK_utils
The file network.py contains an alternative definition of the usual Linear and ConvNd layers using the parametrization described in the article https://arxiv.org/abs/1806.07572. This parametrization gives a consistent scaling behaviour as one increases the width of the network (the number of neurons in the hidden layers). As a result of the reparametrization, we suggest a learning rate around 1.0.

Furthermore network.py contains a module LinearNet which defines a fully-connected network given a list of the number of neurons in each layer. From this module, one can directly calculate the Neural Tangent Kernel, and the activation kernels as described in the article.

Owner

  • Name: Benjamin Keene
  • Login: benkeene
  • Kind: user

Mathematics PhD student at UCF

GitHub Events

Total
Last Year