The library fastFM is an academic project. The time and resources spent developing fastFM are therefore justified by the number of citations of the software. If you publish scientific articles using fastFM, please cite the following article (bibtex entry citation.bib).
Bayer, I. "fastFM: A Library for Factorization Machines" Journal of Machine Learning Research 17, pp. 1-5 (2016)
This repository allows you to use Factorization Machines in Python (2.7 & 3.x) with the well known scikit-learn API. All performance critical code has been written in C and wrapped with Cython. fastFM provides stochastic gradient descent (SGD) and coordinate descent (CD) optimization routines as well as Markov Chain Monte Carlo (MCMC) for Bayesian inference. The solvers can be used for regression, classification and ranking problems. Detailed usage instructions can be found in the online documentation and on arXiv.
fastFM has a continuous integration / testing servers (Travis) for Linux (Ubuntu 14.04 LTS) and OS X Mavericks. Other OSs are not actively supported.
from fastFM import als
fm = als.FMRegression(n_iter=1000, init_stdev=0.1, rank=2, l2_reg_w=0.1, l2_reg_V=0.5)
fm.fit(X_train, y_train)
y_pred = fm.predict(X_test)
Tutorials and other information are available here. The C code is available as subrepository and provides a stand alone command line interface. If you still have questions after reading the documentation please open an issue at GitHub.
| Task | Solver | Loss |
|---|---|---|
| Regression | als, mcmc, sgd | Square Loss |
| Classification | als, mcmc, sgd | Probit(Map), Probit, Sigmoid |
| Ranking | sgd | BPR |
Supported solvers and tasks
binary install (64bit only)
pip install fastFM
source install
Please make sure, that Python and OS bit version agree, e.g. 32bit Python on 64bit OS won't work.
# Install cblas and python-dev header (Linux only).
# - cblas can be installed with libatlas-base-dev or libopenblas-dev (Ubuntu)
$ sudo apt-get install python-dev libopenblas-dev
# Clone the repo including submodules (or clone + `git submodule update --init --recursive`)
$ git clone --recursive https://github.com/ibayer/fastFM.git
# Enter the root directory
$ cd fastFM
# Install Python dependencies (Cython>=0.22, numpy, pandas, scipy, scikit-learn)
$ pip install -r ./requirements.txt
# Compile the C extension.
$ make # build with default python version (python)
$ PYTHON=python3 make # build with custom python version (python3)
# Install fastFM
$ pip install .
The Python tests (pip install nose) can be run with:
nosetests fastFM/fastFM/tests
Please refer to the fastFM-core README for instruction on how to run the
C tests at fastFM/fastFM-core/src/tests.
- Star this repository: keeps contributors motivated
- Open an issue: report bugs or suggest improvements
- Fix errors in the documentation: small changes matter
- Contribute code
Contributions are very welcome! Since this project lives on GitHub we recommend to open a pull request (PR) for code contributions as early as possible. This is the fastest way to get feedback and allows Travis CI to run checks on your changes.
Most information you need to setup your development environment can be learned by adapting the great instructions on https://github.com/scikit-learn/scikit-learn/blob/master/CONTRIBUTING.md . Please ensure that your contribution conforms to the PEP8 Coding Style and includes unit tests where appropriate. More valuable guidelines that apply to fastFM can be found at http://scikit-learn.org/stable/developers/contributing.html#coding-guidelines .
Contributors