Joint Optimization of Cascade Ranking Models

Implementation of the cascade models used in the paper Joint Optimization of Cascade Ranking Models, WSDM 19.

L. Gallagher and R-C. Chen and R. Blanco and J. S. Culpepper. 2019. Joint Optimization of Cascade Ranking Models. In Proc. WSDM. 15-23. DOI: https://doi.org/10.1145/3289600.3290986

Reproduce Instructions

1. Make sure you have installed the dependencies:

   • g++ or clang++
   • cmake 2.8+
   • Python 3.3+
   • MSLR-WEB10K
   • Yahoo LTR Set 1

2. Clone the repo:

   $ git clone https://github.com/rmit-ir/joint-cascade-ranking
   $ cd joint-cascade-ranking
   $ git submodule update --init --recursive

3. Build/Install Dependencies

   Build CEGB (note CEGB was added to LightGBM core after the publication of this work):

   $ cd ext/cegb
   $ mkdir build
   $ cd build
   $ cmake ..
   $ make
   $ cd ../../..

   Build evaluation tools:

   $ make -C eval

   Install Python dependencies:

   $ pip install -r requirements.txt
   $ pip install -e ext/cegb/python-package

4. Configure Datasets

   Create a local.mk file with the path to MSLR and Yahoo datasets:

   $ >local.mk
   $ echo "YAHOO_PATH=/path/to/yahoo" >> local.mk
   $ echo "MSLR_PATH=/path/to/mslr" >> local.mk

   Link to datasets and create qrels:

   $ make -C exp/yahoo
   $ make -C exp/mslr

5. Run baselines

   Yahoo:

   $ ./exp/yahoo/baseline-gbrt.sh
   $ ./exp/yahoo/baseline-cegb.sh

   MSLR:

   $ ./exp/mslr/baseline-gbrt.sh
   $ ./exp/mslr/baseline-cegb.sh

6. Run joint cascade

   $ ./exp/yahoo/reproduce.sh
   $ ./exp/mslr/reproduce.sh