/TransCF

Collaborative Translational Metric Learning (ICDM 2018)

Primary LanguageJupyter Notebook

Collaborative Translational Metric Learning (TransCF)

Overview

Recently, matrix factorization-based recommendation methods have been criticized for the problem raised by the triangle inequality violation. Although several metric learning-based approaches have been proposed to overcome this issue, existing approaches typically project each user to a single point in the metric space, and thus do not suffice to properly model the intensity and the heterogeneity of user-item relationships of implicit feedback. In this paper, we propose TransCF to discover such latent user-item relationships embodied in implicit user-item interactions. Inspired by the translation mechanism popularized by knowledge graph embedding, we construct user-item specific translation vectors by employing the neighborhood information of users and items, and translate each user towards items regarding the user's relationships with the items.

Paper

  • Translational Collaborative Metric Learning (ICDM 2018)

Requirements

How to Run

git clone https://github.com/pcy1302/TransCF.git
cd TransCF
python main.py --recommender TransCF --dataset delicious --lRate 0.005 --mode Val

Configuration

You can evaluate TransCF with different settings. Below is a description of all the configurable parameters:

  • --recommender : 'Choose a recommender.'
  • --dataset : 'Choose a dataset.' (delicious, bookcrossing, ciao, cellphone)
  • --embedding_dim : 'Number of embedding dimensions.'
  • --lRate : 'Learning rate.'
  • --margin' : 'Margin.'
  • --reg1' : 'Distance regularizer.'
  • --reg2' : 'Neighborhood regularizer.'
  • --mode : 'Validation or Test' (Val, Test)
  • --numEpoch : 'Number of epochs.'
  • --num_negatives' : 'Number of negative samples.
  • --batch_size' : 'Batch size.'
  • --rand_seed' : 'Random seed.'
  • --cuda : 'Speficy GPU number'
  • --early_stop : 'Early stop iteration.'

Reproducing the qualitative experiments