LightGBM.jl

LightGBM.jl provides a high-performance Julia interface for Microsoft's LightGBM.

The package adds a couple of convenience features:

  • Automated cross-validation
  • Exhaustive grid search search procedure
  • Integration with MLJ (which also provides the above via different interfaces)

Additionally, the package automatically converts all LightGBM parameters that refer to indices (e.g. categorical_feature) from Julia's one-based indices to C's zero-based indices.

A majority of the C-interfaces are implemented. A few are known to be missing and are tracked.

All major operating systems (Windows, Linux, and Mac OS X) are supported. Julia versions 1.0+ are supported.

Table of Contents

Installation

Please ensure your system meets the pre-requisites for LightGBM. This generally means ensuring that libomp is installed and linkable on your system. See here for Microsoft's installation guide.

Please note that the package actually downloads a precompiled binary so you do not need to install LightGBM first. This is done as a user convenience, and support will be added for supplying ones own LightGBM binary (for GPU acceleration, etc).

To add the package to Julia:

Pkg.add("LightGBM")

Running tests for the package requires the use of the LightGBM example files, download and extract the LightGBM source and set the enviroment variable LIGHTGBM_EXAMPLES_PATH to the root of the source installation. Then you can run the tests by simply doing

Pkg.test("LightGBM")

A simple example using LightGBM example files

First, download LightGBM source and untar it somewhere.

cd ~
wget https://github.com/microsoft/LightGBM/archive/v2.3.1.tar.gz
tar -xf v2.3.1.tar.gz
using LightGBM
using DelimitedFiles

LIGHTGBM_SOURCE = abspath("~/LightGBM-2.3.1")

# Load LightGBM's binary classification example.
binary_test = readdlm(joinpath(LIGHTGBM_SOURCE, "examples", "binary_classification", "binary.test"), '\t')
binary_train = readdlm(joinpath(LIGHTGBM_SOURCE, "examples", "binary_classification", "binary.train"), '\t')
X_train = binary_train[:, 2:end]
y_train = binary_train[:, 1]
X_test = binary_test[:, 2:end]
y_test = binary_test[:, 1]

# Create an estimator with the desired parameters—leave other parameters at the default values.
estimator = LGBMClassification(
    objective = "binary",
    num_iterations = 100,
    learning_rate = .1,
    early_stopping_round = 5,
    feature_fraction = .8,
    bagging_fraction = .9,
    bagging_freq = 1,
    num_leaves = 1000,
    num_class = 1,
    metric = ["auc", "binary_logloss"]
)

# Fit the estimator on the training data and return its scores for the test data.
fit!(estimator, X_train, y_train, (X_test, y_test))

# Predict arbitrary data with the estimator.
predict(estimator, X_train)

# Cross-validate using a two-fold cross-validation iterable providing training indices.
splits = (collect(1:3500), collect(3501:7000))
cv(estimator, X_train, y_train, splits)

# Exhaustive search on an iterable containing all combinations of learning_rate ∈ {.1, .2} and
# bagging_fraction ∈ {.8, .9}
params = [Dict(:learning_rate => learning_rate,
               :bagging_fraction => bagging_fraction) for
          learning_rate in (.1, .2),
          bagging_fraction in (.8, .9)]
search_cv(estimator, X_train, y_train, splits, params)

# Save and load the fitted model.
filename = pwd() * "/finished.model"
savemodel(estimator, filename)
loadmodel!(estimator, filename)

Parameters

Note that a lot of parameters used within this module and in the code and examples are exact matches with those from LightGBM. Not all of these are necessarily supported but see the guide for detailed explanations of what these parameters do and their valid values.

MLJ Support

This package has an interface to MLJ. Exhaustive MLJ documentation is out of scope for here, however the main things are:

The MLJ interface models are

LightGBM.MLJInterface.LGBMClassifier
LightGBM.MLJInterface.LGBMRegressor

And these have the same interface parameters as the estimators

The interface models are generally passed to MLJBase.fit or MLJBase.machine and integrated as part of a larger MLJ pipeline. An example is provided

Custom LightGBM binaries

Though this package comes with a precompiled binary (lib_lightgbm.so for linux, lib_lightgbm.dylib for macos, lib_lightgbm.dll for windows, refer to Microsoft's LightGBM release page), a custom binary can be used with this package (we use Libdl.dlopen to do this). In order to do so, either:

  • Add the directory of your custom binary to the Libdl.DL_LOAD_PATH before calling import LightGBM, e.g. ``` import Libdl push!(Libdl.DLLOADPATH, "/path/to/your/lib_lightgbm/directory")

    import LightGBM ... ```

  • Specify the directory of your custom binary in the environment variables LD_LIBRARY_PATH (for linux), DYLD_LIBRARY_PATH (macos), PATH (windows), or place the custom binary file in the system search path

Note: Libdl.DL_LOAD_PATH will be first searched and used, then the system library paths. If no binaries are found, the program will fallback to using the precompiled binary

Contributors ✨

The list of our Contributors can be found here. Please don't hesitate to add yourself when you contribute.