lorenzmuller/kernelnet_movielens

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 1.13)

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)

kernelnet_movielens's Issues

Is this based on item-based Autoencoder method?

Thx for your sharing.
I'm new to KernelNet, what are the differences between kernelnet and FCN?