Describir: Hyper-parameter optimization for support vector machines using stochastic gradient descent and dual coordinate descent