def __init__(self,
data,
depth,
n_trees,
projection_sparsity='auto',
shape=None,
mmap=False):
"""
Initializes an MRPT index object.
:param data: Input data either as a NxDim numpy ndarray or as a filepath to a binary file containing the data
:param depth: The depth of the trees
:param n_trees: The number of trees used in the index
:param projection_sparsity: Expected ratio of non-zero components in a projection matrix
:param shape: Shape of the data as a tuple (N, dim). Needs to be specified only if loading the data from a file.
:param mmap: If true, the data is mapped into memory. Has effect only if the data is loaded from a file.
:return:
"""
if isinstance(data, np.ndarray):
if len(data) == 0 or len(data.shape) != 2:
raise ValueError(
"The data matrix should be non-empty and two-dimensional")
if data.dtype != np.float32:
raise ValueError("The data matrix should have type float32")
if not data.flags['C_CONTIGUOUS'] or not data.flags['ALIGNED']:
raise ValueError(
"The data matrix has to be C_CONTIGUOUS and ALIGNED")
n_samples, dim = data.shape
elif isinstance(data, str):
if not isinstance(shape, tuple) or len(shape) != 2:
raise ValueError(
"You must specify the shape of the data as a tuple (N, dim) "
"when loading data from a binary file")
n_samples, dim = shape
else:
raise ValueError("Data must be either an ndarray or a filepath")
max_depth = np.ceil(np.log2(n_samples))
if not 1 <= depth <= max_depth:
raise ValueError("Depth should be in range [1, %d]" % max_depth)
if n_trees < 1:
raise ValueError("Number of trees must be positive")
if projection_sparsity == 'auto':
projection_sparsity = 1. / np.sqrt(dim)
elif projection_sparsity is None:
projection_sparsity = 1
elif not 0 < projection_sparsity <= 1:
raise ValueError("Sparsity should be in (0, 1]")
if mmap and os.name == 'nt':
raise ValueError("Memory mapping is not available on Windows")
self.index = mrptlib.MrptIndex(data, n_samples, dim, depth, n_trees,
projection_sparsity, mmap)
self.built = False
def __init__(self, data, shape=None, mmap=False):
"""
Initializes an MRPT index object.
:param data: Input data either as a NxDim numpy ndarray or as a filepath to a binary file containing the data.
:param shape: Shape of the data as a tuple (N, dim). Needs to be specified only if loading the data from a file.
:param mmap: If true, the data is mapped into memory. Has effect only if the data is loaded from a file.
:return:
"""
if isinstance(data, np.ndarray):
if len(data) == 0 or len(data.shape) != 2:
raise ValueError(
"The data matrix should be non-empty and two-dimensional")
if data.dtype != np.float32:
raise ValueError("The data matrix should have type float32")
if not data.flags['C_CONTIGUOUS'] or not data.flags['ALIGNED']:
raise ValueError(
"The data matrix has to be C_CONTIGUOUS and ALIGNED")
n_samples, dim = data.shape
elif isinstance(data, str):
if not isinstance(shape, tuple) or len(shape) != 2:
raise ValueError(
"You must specify the shape of the data as a tuple (N, dim) "
"when loading data from a binary file")
n_samples, dim = shape
elif data is not None:
raise ValueError("Data must be either an ndarray or a filepath")
if mmap and os_name == 'nt':
raise ValueError("Memory mapping is not available on Windows")
if data is not None:
self.index = mrptlib.MrptIndex(data, n_samples, dim, mmap)
self.dim = dim
self.built = False
self.autotuned = False