Batch-Learn is an implementation of ML algorithms which may be applied to on-disk data batch-by-batch, without loading full dataset to memory.
Algorithms included:
- FFM
- NN / MLP
It's extracted from the code written during Outbrain Click Prediction competition on Kaggle and now is undergoing some rewrite and refactoring.
Batch-learn uses CMake as a build tool and depends on following libraries:
- boost-program-options
- boost-iostreams
To compile code you need to install boost libraries and then call:
mkdir build
cd build
cmake ..
make
First, you need to convert to batch-learn format:
batch-learn convert -f ffm -b 24 ffm_dataset.txt -O bl_dataset
To train ffm model and make predictions on test dataset:
batch-learn ffm --train tr1 --test te1 --pred pred.txt
You also may specify validation dataset:
batch-learn ffm --train tr1 --test te1 --val va1 --pred pred.txt
To get list of available commands just run:
batch-learn help
To get help about some specific command:
batch-learn help ffm