A short paper describing the library is now available on arXiv http://arxiv.org/abs/1505.00641
This repository relays on sub-repositories just using git clone .. doesn't fetch them.
git clone --recursive ..
Or do the two-step dance if you wish. You need to run git submodule update --init --recursive from within the fastFM-core/ folder in order to clone them as well.
- glib-2.0
- CXSparse (included)
- gsl 1.15-1 (only testsuite)
- argp (included by default in Linux)
install depenencies: -------------------this worked on ubuntu 14.04: sudo apt-get install libglib2.0-dev libatlas-base-dev and for the testsuite also sudo apt-get install libgsl0-dev
Library compiles on OSX (thus python interface works), however console interface doesn't.
Recommended way to manage dependencies is [Homebrew package manager](https://brew.sh). If you have brew installed, dependencies can be installed by running command brew install glib gsl argp-standalone.
It should be possible to compile the library on Windows. I'm developing on linux but have received multiple requests from people who want to run this library on other platforms. Please let me know about issues you ran into or how you manged to compile on other platfroms (or just open a PR) so that we include this information here.
compile cli: make cli you can find the executible in fastFM-core/bin.
demo/data/ contains sample input files for various tasks and examples how to use fastFM's CLI interface can be found in demo/Makefile
- cd
src/tests/ - bulid tests
make all - run all tests
make check - run valgrind memory check on sparse_test.c
make mem_check - run valgrind to check for errors
valgrind -v ./a.out >& out
cli example:
./fastFM-core/bin/fastfm data/train_regression data/test_regression --task regression --init-var=0.11 --n-iter=123 --solver='mcmc' --rank 7 --l2-reg=.22