ZhengHui
/
kernelNet
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
3 Commits
1 Branch
0 Tags
7.3 MiB
 
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'master'
${ noResults }
Go to file
ZhengHui 3555471c65
Update README.md 		3 years ago
ml-1m init 3 years ago
.gitignore Initial commit 3 years ago
LICENSE Initial commit 3 years ago
LICENSE-2.0.txt init 3 years ago
README.md Update README.md 3 years ago
dataLoader.py init 3 years ago
kernelNet_ml1m.py init 3 years ago

README.md

kernelNet MovieLens-1M

State of the art model for MovieLens-1M.

This is a minimal implementation of a kernelNet sparsified autoencoder for MovieLens-1M. See http://proceedings.mlr.press/v80/muller18a.html

Setup

Download this repository

Requirements

  • numpy
  • scipy
  • tensorflow (tested with version 1.4.1)

Dataset

Expects MovieLens-1M dataset in a subdirectory named ml-1m. Get it here https://grouplens.org/datasets/movielens/1m/

or on linux run in the project directory

wget --output-document=ml-1m.zip http://www.grouplens.org/system/files/ml-1m.zip; unzip ml-1m.zip

Run

python kernelNet_ml1m.py optional arguments are the L2 and sparsity regularization strength. Default is 60. and 0.013

Results

with the default parameters this slightly outperforms the paper model at 0.823 validation RMSE (10-times repeated random sub-sampling validation)