Skip to content
Codes and experiments for paper "Automated Spectral Kernel Learning". Preprint.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Automated Spectral Kernel Learning


This repository provides the code used to run the experiments of the paper "Automated Spectral Kernel Learning" (


  • Python 3.7.4
  • Pytorch 1.2.0
  • CUDA 10.1.168
  • cuDnn 7.6.0
  • GPU: Tesla P100 16GB

Core functions

  • implements the algorithm to construct an one-layer neural network, including initialization of trainable weights and untrainable biases as well as feature mapping (cosine as activation).
  • implements useful tools including load svmlight style dataset and classic datasets used in Pytorch but also various loss functions are introduced.
  • records optimal parameters for the proposed algorithm.


  1. Download datasets for multi-class classification (
  2. Run the script to tune parameters and record them in
  1. Run the script to obtain results in Experiment section
You can’t perform that action at this time.