def _build_k_prototypes(self, nprot, data):
# K-Means / K-Medoids algorithm
# should return a N-length array with he indices of the chosen data
grouper = K_Means()
grouper._centroid_mode = "index"
grouper.fit(data,
nprot,
iters=300,
mode="cuda",
cuda_mem='manual',
tol=1e-4,
max_iters=300)
centroids = grouper.centroids
nclusters = centroids.shape[0]
# TODO - very inefficient
k_labels = np.zeros(nclusters, dtype=np.int32)
for k in xrange(nclusters):
dist = data - centroids[k]
dist = dist**2
dist = dist.sum(axis=1)
k_labels[k] = dist.argmin()
return k_labels
from sklearn.cluster import KMeans from numbapro import cuda import sys n_samples = 1e4 n_features = 2 centers = 6 n_samples = np.int(n_samples) data, gt = make_blobs(n_samples = n_samples, n_features = n_features, centers = centers) data = data.astype(np.float32) n_samples_sqrt = np.int(np.sqrt(n_samples)) n_clusters = [n_samples_sqrt / 2, n_samples_sqrt] n_clusters = map(int,n_clusters) generator = K_Means(cuda_mem="manual") generator.n_clusters = 176 generator.max_iter = 1 generator.N = n_samples generator.D = n_features generator.centroids = generator._init_centroids(data) generator.labels = cuda.pinned_array(shape=generator.N, dtype = np.int32) generator._dists = cuda.pinned_array(shape=generator.N, dtype = np.float32) generator._compute_cuda_dims(data) gridDim = generator._gridDim blockDim = generator._blockDim
n_samples = 1e5 n_features = 100 n_samples = np.int(n_samples) data, gt = make_blobs(n_samples = n_samples, n_features = n_features, centers = centers) data = data.astype(np.float32) n_samples_sqrt = np.int(np.sqrt(n_samples)) n_clusters = [n_samples_sqrt / 2, n_samples_sqrt] n_clusters = map(int,n_clusters) # generator = K_Means(cuda_mem="manual") # generator.n_clusters = n_clusters[-1] # generator.max_iter = 3 # generator.fit(data) generator = K_Means(cuda_mem="manual") generator.N = n_samples generator.D = n_features generator.n_clusters = n_clusters[-1] generator.centroids = generator._init_centroids(data) generator.labels = np.empty(shape=generator.N, dtype = np.int32) generator._dists = np.empty(shape=generator.N, dtype = np.float32) generator._compute_cuda_dims(data) gridDim = generator._gridDim blockDim = generator._blockDim dData = cuda.to_device(data) dCentroids = cuda.to_device(generator.centroids)
n_features = 100
n_samples = np.int(n_samples)
data, gt = make_blobs(n_samples=n_samples,
n_features=n_features,
centers=centers)
data = data.astype(np.float32)
n_samples_sqrt = np.int(np.sqrt(n_samples))
n_clusters = [n_samples_sqrt / 2, n_samples_sqrt]
n_clusters = map(int, n_clusters)
# generator = K_Means(cuda_mem="manual")
# generator.n_clusters = n_clusters[-1]
# generator.max_iter = 3
# generator.fit(data)
generator = K_Means(cuda_mem="manual")
generator.N = n_samples
generator.D = n_features
generator.n_clusters = n_clusters[-1]
generator.centroids = generator._init_centroids(data)
generator.labels = np.empty(shape=generator.N, dtype=np.int32)
generator._dists = np.empty(shape=generator.N, dtype=np.float32)
generator._compute_cuda_dims(data)
gridDim = generator._gridDim
blockDim = generator._blockDim
dData = cuda.to_device(data)
dCentroids = cuda.to_device(generator.centroids)