Some machine learning kernels write in CUDA with python binding.
Implements K-Means clustering algorithm with CUDA, and support Numpy interface.
import numpy as np
from PyCUDAML.KMeans import KMeans
# Design matrix as a numpy array.
X = np.random.rand(60000, 748)
k = KMeans(num_clusters=16, threshold=1e-2,
seed=1117, max_iter=1000)
# Returned cluster centers are also numpy array.
cluster_centers = k.fit(X)You need to install CUDA first.
python setup.py install