Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

NeurIPS'17 DropoutNet: Addressing Cold Start in Recommender Systems

Authors: Maksims Volkovs, Guangwei Yu, Tomi Poutanen

UPDATE: We added TensorFlow2 and PyTorch implimentations!


This repository contains full implementation of the DropoutNet model and includes both training and evaluation routines. We also provide the ACM RecSys 2017 Challenge dataset that we further split into three subsets for warm start, user cold start and item cold start evaluation. The aim is to train a single model that can be applied to all three tasks and we report validation accuracy on each task during training.

Furthermore per request, we also provide scripts and all necessary data to run the Citeulike cold-start experiment. See section on Citeulike below for further details as well as links to the packaged data.


To run the model, download the dataset from here. With this dataset we have also included pre-trained Weighted Factorization model (WMF)[Hu et al., 2008], that is used as preference input to the DropoutNet. WMF produces competitive performance on warm start but doesn't generalize to cold start. So this code demonstrates how to apply DropoutNet to provide cold start capability to WMF. The format of the data is as follows:				
└─ eval					// use path to this folder in --data-dir
   ├─ trained				// WMF model
   │  └─ warm				
   │     ├─ U.csv.bin			// numpy binarized WMF user preference latent vectors (U)
   │     └─ V.csv.bin			// numpy binarized WMF item preference latent vectors (V)
   ├─ warm				
   │  ├─ test_cold_item.csv		// validation interactions for item cold start 
   │  ├─ test_cold_item_item_ids.csv	// targets item ids for item cold start
   │  ├─ test_cold_user.csv    		// validation interactions for user cold start
   │  ├─ test_cold_user_item_ids.csv	// target user ids for user cold start
   │  ├─ test_warm.csv			// validation interactions for warm start
   │  ├─ test_warm_item_ids.csv		// target item ids for warm start
   │  └─ train.csv			// training interactions
   ├─ item_features_0based.txt		// item features in libsvm format
   └─ user_features_0based.txt		// user features in libsvm format
interactions are stored in csv as:
where INTERACTION_TYPE is one of:
  0: impression
  1: click
  2: bookmark
  3: reply
  5: recruiter interest

Running training code

  1. Download the dataset, extract and keep the directory structure.

  2. run

    • for usage, run with --help
    • default setting trains a two layer neural network with hyperparameters selected for the RecSys data
    • gpu is used for training by default and cpu for inference
  3. (Optionally) launch tensorboard to monitor progress by tensorboard --logdir=<log_path>

During training recall@50,100,...,500 accuracy is shown every 50K updates for warm start, user cold start and item cold start validation sets.


  • Make sure --data-dir points to the eval/ folder, not the root
  • On our environment (described above) 50K updates takes approximately 14 minutes with the default GPU/CPU setting.
  • By default, training happens on GPU while inference and batch generation is on CPU.

Validation Curves


In addition to Recsys, we also provide pipeline to run the publicly available Citeulike data. Note that, as mentioned in the paper, we evaluate cold start the same way as the CTR paper while the warm start evaluation is modified. For convenience, we have proivded our evaluation split for both cold and warm start, item features, as well as the WMF user item preference latent vectors available here.

The citeulike warm and cold models are trained separately as their validation sets differ. Please use the scripts and to run the experiments on the Citeulike dataset.

Point --data-dir to your extracted eval folder after extracting citeu.tar.gz. Sample training runs with respective validation performance are shown below per 1000 updates.


Code for the NeurIPS'17 paper "DropoutNet: Addressing Cold Start in Recommender Systems"







No releases published


No packages published