ThunderSVM Ruby

ThunderSVM - high performance parallel SVMs - for Ruby

🔥 Uses GPUs and multi-core CPUs for blazing performance

For a great intro on support vector machines, check out this video.

Installation

Add this line to your application’s Gemfile:

gem "thundersvm"

On Mac, also install OpenMP:

brew install libomp

Getting Started

Prep your data

x = [[1, 2], [3, 4], [5, 6], [7, 8]]
y = [1, 2, 3, 4]

Train a model

model = ThunderSVM::Regressor.new
model.fit(x, y)

Use ThunderSVM::Classifier for classification and ThunderSVM::Model for other models

Make predictions

model.predict(x)

Save the model to a file

model.save_model("model.txt")

Load the model from a file

model = ThunderSVM.load_model("model.txt")

Get support vectors

model.support_vectors

Cross-Validation

Perform cross-validation

model.cv(x, y)

Specify the number of folds

model.cv(x, y, folds: 5)

Parameters

Defaults shown below

ThunderSVM::Model.new(
  svm_type: :c_svc,    # type of SVM (c_svc, nu_svc, one_class, epsilon_svr, nu_svr)
  kernel: :rbf,        # type of kernel function (linear, polynomial, rbf, sigmoid)
  degree: 3,           # degree in kernel function
  gamma: nil,          # gamma in kernel function
  coef0: 0,            # coef0 in kernel function
  c: 1,                # parameter C of C-SVC, epsilon-SVR, and nu-SVR
  nu: 0.5,             # parameter nu of nu-SVC, one-class SVM, and nu-SVR
  epsilon: 0.1,        # epsilon in loss function of epsilon-SVR
  max_memory: 8192,    # constrain the maximum memory size (MB) that thundersvm uses
  tolerance: 0.001,    # tolerance of termination criterion
  probability: false,  # whether to train a SVC or SVR model for probability estimates
  gpu: 0,              # specify which gpu to use
  cores: nil,          # number of cpu cores to use (defaults to all)
  verbose: false       # verbose mode
)

Data

Data can be a Ruby array

[[1, 2], [3, 4], [5, 6], [7, 8]]

Or a Numo array

Numo::DFloat.cast([[1, 2], [3, 4], [5, 6], [7, 8]])

Or the path a file in libsvm format (better for sparse data)

model.fit("train.txt")
model.predict("test.txt")

GPUs

To run ThunderSVM on GPUs, you’ll need to build the library from source.

Linux

git clone --recursive --branch v0.3.4 https://github.com/Xtra-Computing/thundersvm
cd thundersvm
mkdir build
cd build
cmake ..
make

Specify the path to the shared library with:

ThunderSVM.ffi_lib = "path/to/build/lib/libthundersvm.so"

Windows

Follow the official instructions. Specify the path to the shared library with:

ThunderSVM.ffi_lib = "path/to/build/lib/libthundersvm.dll"

Resources

• ThunderSVM: A Fast SVM Library on GPUs and CPUs
• A Practical Guide to Support Vector Classification

History

View the changelog

Contributing

Everyone is encouraged to help improve this project. Here are a few ways you can help:

• Report bugs
• Fix bugs and submit pull requests
• Write, clarify, or fix documentation
• Suggest or add new features

To get started with development:

git clone https://github.com/ankane/thundersvm-ruby.git
cd thundersvm-ruby
bundle install
bundle exec rake test