Beta release EMF and other baselines are implemented with TensorFlow v1.
We recommend users to install required packages dependencies for the EMF models and experiments, using Conda. To install Python and other dependencies, which you can do directly using the provided environment.yml file
:
conda env create -f environment.yml
source activate EMF
The emf
Conda environment can be used to run any number of experiments and examples included in this project/repository. We note that experiments and data download for other parts of this project are implemented and configured with Snakemake which will be installed as part of the EMF
environment.
To use GPUs, install the emf-gpu
environment via:
conda env create -f environment-gpu.yml
source activate EMF-gpu
See experiments/all-mf-monte-carlo-cv
.
Hyperparameter searches for different MF models are implemented and can be run using snakemake
in the following directories in experiments/
:
single-species-mf-hp-search
- MF, KPMF, MF with bias, KPMF with bias, NGMC models.xsmf-hp-search
- XSMF modelk-xsmf-hp-search
- Kernelized XSMF modelk-xsmf-b-hp-search
- Kernelized XSMF with bias model