Hashed Factorization Machine with Follow The Regularized Leader for Kaggle Avazu Click-Through Rate Competition
Python
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 1 commit behind comadan:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
FM_FTRL_machine.py
README.md
runmodel_dropout_example.py
runmodel_example.py

README.md

FM_FTRL

Hashed Factorization Machine with Follow The Regularized Leader online learning for Kaggle Avazu Click-Through Rate Competition

Based on Tinrtgu's code at: http://www.kaggle.com/c/avazu-ctr-prediction/forums/t/10927/beat-the-benchmark-with-less-than-1mb-of-memory

Features:

  • online learning
  • feature hashing to limit memory footprint
  • Follow the Regularized Leader (FTRL) optimization: http://research.google.com/pubs/pub41159.html
  • Factorization Machine (low rank matrix factorization)
  • L1 and L2 regularization
  • dropout regularization option

Run fast with pypy:

pypy runmodel_example.py

or run slow with regular python:

python runmodel_example.py