def test_percentile():
d = da.from_array(cupy.ones((16,)), chunks=(4,))
qs = np.array([0, 50, 100])
assert_eq(
da.percentile(d, qs, interpolation="midpoint"),
np.array([1, 1, 1], dtype=d.dtype),
)
x = cupy.array([0, 0, 5, 5, 5, 5, 20, 20])
d = da.from_array(x, chunks=(3,))
result = da.percentile(d, qs, interpolation="midpoint")
assert_eq(result, np.array([0, 5, 20], dtype=result.dtype))
# Currently fails, tokenize(cupy.array(...)) is not deterministic.
# See https://github.com/dask/dask/issues/6718
# assert same_keys(
# da.percentile(d, qs),
# da.percentile(d, qs)
# )
assert not same_keys(
da.percentile(d, qs, interpolation="midpoint"),
da.percentile(d, [0, 50], interpolation="midpoint"),
)
def test_unknown_chunk_sizes(method):
x = da.random.random(1000, chunks=(100,))
x._chunks = ((np.nan,) * 10,)
result = da.percentile(x, 50, method=method).compute()
assert 0.1 < result < 0.9
a, b = da.percentile(x, [40, 60], method=method).compute()
assert 0.1 < a < 0.9
assert 0.1 < b < 0.9
assert a < b
def test_unknown_chunk_sizes():
x = da.random.random(1000, chunks=(100, ))
x._chunks = ((np.nan, ) * 10, )
result = da.percentile(x, 50).compute()
assert 0.1 < result < 0.9
a, b = da.percentile(x, [40, 60]).compute()
assert 0.1 < a < 0.9
assert 0.1 < b < 0.9
assert a < b
def _calculate_summary_statistics(self):
data = self._lazy_data()
_raveled = data.ravel()
_mean, _std, _min, _q1, _q2, _q3, _max = da.compute(
da.nanmean(data),
da.nanstd(data),
da.nanmin(data),
da.percentile(_raveled, [25, ]),
da.percentile(_raveled, [50, ]),
da.percentile(_raveled, [75, ]),
da.nanmax(data), )
return _mean, _std, _min, _q1, _q2, _q3, _max
def test_percentile():
d = da.ones((16,), chunks=(4,))
assert eq(da.percentile(d, [0, 50, 100]), [1, 1, 1])
x = np.array([0, 0, 5, 5, 5, 5, 20, 20])
d = da.from_array(x, chunks=(3,))
assert eq(da.percentile(d, [0, 50, 100]), [0, 5, 20])
x = np.array(['a', 'a', 'd', 'd', 'd', 'e'])
d = da.from_array(x, chunks=(3,))
assert eq(da.percentile(d, [0, 50, 100]), ['a', 'd', 'e'])
def _calculate_summary_statistics(self):
data = self._lazy_data()
_raveled = data.ravel()
_mean, _std, _min, _q1, _q2, _q3, _max = da.compute(
da.nanmean(data),
da.nanstd(data),
da.nanmin(data),
da.percentile(_raveled, [25, ]),
da.percentile(_raveled, [50, ]),
da.percentile(_raveled, [75, ]),
da.nanmax(data), )
return _mean, _std, _min, _q1, _q2, _q3, _max
def test_percentile():
d = da.ones((16,), blockshape=(4,))
assert eq(da.percentile(d, [0, 50, 100]), [1, 1, 1])
x = np.array([0, 0, 5, 5, 5, 5, 20, 20])
d = da.from_array(x, blockshape=(3,))
assert eq(da.percentile(d, [0, 50, 100]), [0, 5, 20])
x = np.array(['a', 'a', 'd', 'd', 'd', 'e'])
d = da.from_array(x, blockshape=(3,))
assert eq(da.percentile(d, [0, 50, 100]), ['a', 'd', 'e'])
def test_percentiles_with_unknown_chunk_sizes():
rs = da.random.RandomState(RandomState=cupy.random.RandomState)
x = rs.random(1000, chunks=(100,))
x._chunks = ((np.nan,) * 10,)
result = da.percentile(x, 50, interpolation="midpoint").compute()
assert type(result) == cupy.core.core.ndarray
assert 0.1 < result < 0.9
a, b = da.percentile(x, [40, 60], interpolation="midpoint").compute()
assert type(a) == cupy.core.core.ndarray
assert type(b) == cupy.core.core.ndarray
assert 0.1 < a < 0.9
assert 0.1 < b < 0.9
assert a < b
def test_percentiles_with_empty_arrays():
x = da.from_array(cupy.ones(10), chunks=((5, 0, 5), ))
res = da.percentile(x, [10, 50, 90], interpolation="midpoint")
assert type(res._meta) == cupy.ndarray
assert_eq(res, res) # Check that _meta and computed arrays match types
assert_eq(res, np.array([1, 1, 1], dtype=x.dtype), check_type=False)
def _calculate_summary_statistics(self, rechunk=True):
if rechunk is True:
# Use dask auto rechunk instead of HyperSpy's one, what should be
# better for these operations
rechunk = "dask_auto"
data = self._lazy_data(rechunk=rechunk)
_raveled = data.ravel()
_mean, _std, _min, _q1, _q2, _q3, _max = da.compute(
da.nanmean(data),
da.nanstd(data),
da.nanmin(data),
da.percentile(_raveled, [25, ]),
da.percentile(_raveled, [50, ]),
da.percentile(_raveled, [75, ]),
da.nanmax(data), )
return _mean, _std, _min, _q1, _q2, _q3, _max
def test_percentile_with_categoricals():
try:
import pandas as pd
except ImportError:
return
x0 = pd.Categorical(["Alice", "Bob", "Charlie", "Dennis", "Alice", "Alice"])
x1 = pd.Categorical(["Alice", "Bob", "Charlie", "Dennis", "Alice", "Alice"])
dsk = {("x", 0): x0, ("x", 1): x1}
x = da.Array(dsk, "x", chunks=((6, 6),))
p = da.percentile(x, [50])
assert (p.compute().categories == x0.categories).all()
assert (p.compute().codes == [0]).all()
assert same_keys(da.percentile(x, [50]), da.percentile(x, [50]))
def fit(
self,
X: Union[ArrayLike, DataFrameType],
y: Optional[Union[ArrayLike, SeriesType]] = None,
) -> "RobustScaler":
q_min, q_max = self.quantile_range
if not 0 <= q_min <= q_max <= 100:
raise ValueError("Invalid quantile range: %s" %
str(self.quantile_range))
if isinstance(X, dd.DataFrame):
n_columns = len(X.columns)
partition_lengths = X.map_partitions(len).compute()
dtype = np.find_common_type(X.dtypes, [])
blocks = X.to_delayed()
X = da.vstack([
da.from_delayed(block.values,
shape=(length, n_columns),
dtype=dtype)
for block, length in zip(blocks, partition_lengths)
])
quantiles: Any = [
da.percentile(col, [q_min, 50.0, q_max]) for col in X.T
]
quantiles = da.vstack(quantiles).compute()
self.center_: List[float] = quantiles[:, 1]
self.scale_: List[float] = quantiles[:, 2] - quantiles[:, 0]
self.scale_ = _handle_zeros_in_scale(self.scale_, copy=False)
self.n_features_in_: int = X.shape[1]
return self
def _calculate_summary_statistics(self, rechunk=True):
if rechunk is True:
# Use dask auto rechunk instead of HyperSpy's one, what should be
# better for these operations
rechunk = "dask_auto"
data = self._lazy_data(rechunk=rechunk)
_raveled = data.ravel()
_mean, _std, _min, _q1, _q2, _q3, _max = da.compute(
da.nanmean(data),
da.nanstd(data),
da.nanmin(data),
da.percentile(_raveled, [25, ]),
da.percentile(_raveled, [50, ]),
da.percentile(_raveled, [75, ]),
da.nanmax(data), )
return _mean, _std, _min, _q1, _q2, _q3, _max
def fit(self, X, y=None):
q_min, q_max = self.quantile_range
if not 0 <= q_min <= q_max <= 100:
raise ValueError("Invalid quantile range: %s" % str(self.quantile_range))
if isinstance(X, dd.DataFrame):
n_columns = len(X.columns)
partition_lengths = X.map_partitions(len).compute()
dtype = np.find_common_type(X.dtypes, [])
blocks = X.to_delayed()
X = da.vstack(
[
da.from_delayed(
block.values, shape=(length, n_columns), dtype=dtype
)
for block, length in zip(blocks, partition_lengths)
]
)
quantiles = [da.percentile(col, [q_min, 50., q_max]) for col in X.T]
quantiles = da.vstack(quantiles).compute()
self.center_ = quantiles[:, 1]
self.scale_ = quantiles[:, 2] - quantiles[:, 0]
self.scale_ = skdata._handle_zeros_in_scale(self.scale_, copy=False)
return self
def from_bcolz(x, chunksize=None, categorize=True, index=None, **kwargs):
""" Read dask Dataframe from bcolz.ctable
Parameters
----------
x : bcolz.ctable
Input data
chunksize : int (optional)
The size of blocks to pull out from ctable. Ideally as large as can
comfortably fit in memory
categorize : bool (defaults to True)
Automatically categorize all string dtypes
index : string (optional)
Column to make the index
See Also
--------
from_array: more generic function not optimized for bcolz
"""
import dask.array as da
import bcolz
if isinstance(x, (str, unicode)):
x = bcolz.ctable(rootdir=x)
bc_chunklen = max(x[name].chunklen for name in x.names)
if chunksize is None and bc_chunklen > 10000:
chunksize = bc_chunklen
categories = dict()
if categorize:
for name in x.names:
if (np.issubdtype(x.dtype[name], np.string_) or
np.issubdtype(x.dtype[name], np.unicode_) or
np.issubdtype(x.dtype[name], np.object_)):
a = da.from_array(x[name], chunks=(chunksize * len(x.names),))
categories[name] = da.unique(a)
columns = tuple(x.dtype.names)
divisions = (0,) + tuple(range(-1, len(x), chunksize))[1:]
if divisions[-1] != len(x) - 1:
divisions = divisions + (len(x) - 1,)
new_name = 'from_bcolz' + next(tokens)
dsk = dict(((new_name, i),
(dataframe_from_ctable,
x,
(slice(i * chunksize, (i + 1) * chunksize),),
None, categories))
for i in range(0, int(ceil(len(x) / chunksize))))
result = DataFrame(dsk, new_name, columns, divisions)
if index:
assert index in x.names
a = da.from_array(x[index], chunks=(chunksize * len(x.names),))
q = np.linspace(0, 100, len(x) // chunksize + 2)
divisions = da.percentile(a, q).compute()
return set_partition(result, index, divisions, **kwargs)
else:
return result
def test_percentiles_with_empty_q():
x = da.from_array(cupy.ones(10), chunks=((5, 0, 5),))
result = da.percentile(x, [], interpolation="midpoint")
assert type(result._meta) == cupy.core.core.ndarray
assert_eq(result, result) # Check that _meta and computed arrays match types
assert_eq(result, np.array([], dtype=x.dtype))
def from_bcolz(x, chunksize=None, categorize=True, index=None, **kwargs):
""" Read dask Dataframe from bcolz.ctable
Parameters
----------
x : bcolz.ctable
Input data
chunksize : int (optional)
The size of blocks to pull out from ctable. Ideally as large as can
comfortably fit in memory
categorize : bool (defaults to True)
Automatically categorize all string dtypes
index : string (optional)
Column to make the index
See Also
--------
from_array: more generic function not optimized for bcolz
"""
import dask.array as da
import bcolz
if isinstance(x, (str, unicode)):
x = bcolz.ctable(rootdir=x)
bc_chunklen = max(x[name].chunklen for name in x.names)
if chunksize is None and bc_chunklen > 10000:
chunksize = bc_chunklen
categories = dict()
if categorize:
for name in x.names:
if (np.issubdtype(x.dtype[name], np.string_) or
np.issubdtype(x.dtype[name], np.unicode_) or
np.issubdtype(x.dtype[name], np.object_)):
a = da.from_array(x[name], chunks=(chunksize * len(x.names),))
categories[name] = da.unique(a)
columns = tuple(x.dtype.names)
divisions = (0,) + tuple(range(-1, len(x), chunksize))[1:]
if divisions[-1] != len(x) - 1:
divisions = divisions + (len(x) - 1,)
new_name = 'from_bcolz' + next(tokens)
dsk = dict(((new_name, i),
(dataframe_from_ctable,
x,
(slice(i * chunksize, (i + 1) * chunksize),),
None, categories))
for i in range(0, int(ceil(len(x) / chunksize))))
result = DataFrame(dsk, new_name, columns, divisions)
if index:
assert index in x.names
a = da.from_array(x[index], chunks=(chunksize * len(x.names),))
q = np.linspace(0, 100, len(x) // chunksize + 2)
divisions = da.percentile(a, q).compute()
return set_partition(result, index, divisions, **kwargs)
else:
return result
def test_percentile_with_categoricals():
try:
import pandas as pd
except ImportError:
return
x0 = pd.Categorical(['Alice', 'Bob', 'Charlie', 'Dennis', 'Alice', 'Alice'])
x1 = pd.Categorical(['Alice', 'Bob', 'Charlie', 'Dennis', 'Alice', 'Alice'])
dsk = {('x', 0): x0, ('x', 1): x1}
x = da.Array(dsk, 'x', chunks=((6, 6),))
p = da.percentile(x, [50])
assert (p.compute().categories == x0.categories).all()
assert (p.compute().codes == [0]).all()
assert same_keys(da.percentile(x, [50]),
da.percentile(x, [50]))
def test_percentile_with_categoricals():
try:
import pandas as pd
except ImportError:
return
x0 = pd.Categorical(['Alice', 'Bob', 'Charlie', 'Dennis', 'Alice', 'Alice'])
x1 = pd.Categorical(['Alice', 'Bob', 'Charlie', 'Dennis', 'Alice', 'Alice'])
dsk = {('x', 0): x0, ('x', 1): x1}
x = da.Array(dsk, 'x', chunks=((6, 6),))
p = da.percentile(x, [50])
assert (p.compute().categories == x0.categories).all()
assert (p.compute().codes == [0]).all()
assert same_keys(da.percentile(x, [50]),
da.percentile(x, [50]))
def test_percentiles_with_scaler_percentile(internal_method, q):
# Regression test to ensure da.percentile works with scalar percentiles
# See #3020
d = da.ones((16, ), chunks=(4, ))
assert_eq(
da.percentile(d, q, internal_method=internal_method),
np.array([1], dtype=d.dtype),
)
def test_percentile(method):
d = da.ones((16,), chunks=(4,))
qs = [0, 50, 100]
assert_eq(da.percentile(d, qs, method=method), np.array([1, 1, 1], dtype=d.dtype))
x = np.array([0, 0, 5, 5, 5, 5, 20, 20])
d = da.from_array(x, chunks=(3,))
result = da.percentile(d, qs, method=method)
assert_eq(result, np.array([0, 5, 20], dtype=result.dtype))
assert same_keys(
da.percentile(d, qs, method=method), da.percentile(d, qs, method=method)
)
assert not same_keys(
da.percentile(d, qs, method=method), da.percentile(d, [0, 50], method=method)
)
if method != "tdigest":
x = np.array(["a", "a", "d", "d", "d", "e"])
d = da.from_array(x, chunks=(3,))
assert_eq(
da.percentile(d, [0, 50, 100]), np.array(["a", "d", "e"], dtype=x.dtype)
)
def test_percentiles_with_scaler_percentile(q):
# Regression test to ensure da.percentile works with scalar percentiles
# See #3020
d = da.from_array(cupy.ones((16,)), chunks=(4,))
result = da.percentile(d, q, interpolation="midpoint")
assert type(result._meta) == cupy.core.core.ndarray
assert_eq(result, result) # Check that _meta and computed arrays match types
assert_eq(result, np.array([1], dtype=d.dtype))
def test_percentile():
d = da.from_array(cupy.ones((16,)), chunks=(4,))
qs = np.array([0, 50, 100])
assert_eq(
da.percentile(d, qs, interpolation="midpoint"),
np.array([1, 1, 1], dtype=d.dtype),
)
x = cupy.array([0, 0, 5, 5, 5, 5, 20, 20])
d = da.from_array(x, chunks=(3,))
result = da.percentile(d, qs, interpolation="midpoint")
assert_eq(result, np.array([0, 5, 20], dtype=result.dtype))
assert not same_keys(
da.percentile(d, qs, interpolation="midpoint"),
da.percentile(d, [0, 50], interpolation="midpoint"),
)
def _run_dask_numpy_quantile(data, k):
w = 100.0 / k
p = da.arange(w, 100 + w, w)
if p[-1] > 100.0:
p[-1] = 100.0
q = da.percentile(data.flatten(), p)
q = da.unique(q)
return q
def statistics(self, data, pca_stats=None):
# set headers
if pca_stats: # for pca
if pca_stats["eigenvals"] is not None:
self.stats_header.setText("Eigenvalue: {} ({}%)".format(
round(pca_stats["eigenvals"][self.pc_id - 1], 2),
round(pca_stats["eigenvals_%"][self.pc_id - 1], 2)))
self.stats_header.setToolTip(
"It shows how are the dispersion of the data with respect to its component"
)
else:
self.stats_header.setText("Eigenvalue: --")
self.stats_header.setToolTip(
"Is only available when the components are computed with the plugin"
)
else: # for aoi
self.stats_header.setText("Pixels in AOI: {}".format(
round(data.size if data.size > 1 else 0, 2)))
self.stats_header.setToolTip("")
# restore or compute the statistics
if self.QCBox_StatsLayer.currentText(
) == self.pc_name and self.stats_pc is not None:
min, max, std, p25, p50, p75 = self.stats_pc
else:
da_data = da.from_array(data, chunks=(8000000, ))
min = da.min(da_data).compute()
max = da.max(da_data).compute()
std = da.std(da_data).compute()
p25 = da.percentile(da_data, 25).compute()[0]
p50 = da.percentile(da_data, 50).compute()[0]
p75 = da.percentile(da_data, 75).compute()[0]
if self.QCBox_StatsLayer.currentText() == self.pc_name:
self.stats_pc = (min, max, std, p25, p50, p75)
# set in dialog
self.stats_min.setText(str(round(min, 2)))
self.stats_max.setText(str(round(max, 2)))
self.stats_std.setText(str(round(std, 2)))
self.stats_p25.setText(str(round(p25, 2)))
self.stats_p50.setText(str(round(p50, 2)))
self.stats_p75.setText(str(round(p75, 2)))
def dasky_freedman_bin_width(data, return_bins=True):
r"""Dask version of freedman_bin_width
Parameters
----------
data : dask array
the data
return_bins : bool (optional)
if True, then return the bin edges
Returns
-------
width : float
optimal bin width using Scott's rule
bins : ndarray
bin edges: returned if `return_bins` is True
Notes
-----
The optimal bin width is
.. math::
\Delta_b = \frac{2(q_{75} - q_{25})}{n^{1/3}}
where :math:`q_{N}` is the :math:`N` percent quartile of the data, and
:math:`n` is the number of data points.
See Also
--------
knuth_bin_width,
scotts_bin_width,
astroML.plotting.hist
"""
if not isinstance(data, da.Array):
raise TypeError('data has to be a dask array')
if data.ndim != 1:
data = data.flatten()
n = data.size
v25, v75 = da.percentile(data, [25, 75])
dx = 2 * (v75 - v25) * 1. / (n ** (1. / 3))
c_dx, mx, mn = da.compute(dx, data.max(), data.min())
if return_bins:
Nbins = np.ceil((mx - mn) * 1. / c_dx)
Nbins = max(1, Nbins)
bins = mn + c_dx * np.arange(Nbins + 1)
return c_dx, bins
else:
return c_dx
def dasky_freedman_bin_width(data, return_bins=True):
r"""Dask version of freedman_bin_width
Parameters
----------
data : dask array
the data
return_bins : bool (optional)
if True, then return the bin edges
Returns
-------
width : float
optimal bin width using Scott's rule
bins : ndarray
bin edges: returned if `return_bins` is True
Notes
-----
The optimal bin width is
.. math::
\Delta_b = \frac{2(q_{75} - q_{25})}{n^{1/3}}
where :math:`q_{N}` is the :math:`N` percent quartile of the data, and
:math:`n` is the number of data points.
See Also
--------
knuth_bin_width,
scotts_bin_width,
astroML.plotting.hist
"""
if not isinstance(data, da.Array):
raise TypeError('data has to be a dask array')
if data.ndim != 1:
data = data.flatten()
n = data.size
v25, v75 = da.percentile(data, [25, 75])
dx = 2 * (v75 - v25) * 1. / (n**(1. / 3))
c_dx, mx, mn = da.compute(dx, data.max(), data.min())
if return_bins:
Nbins = np.ceil((mx - mn) * 1. / c_dx)
Nbins = max(1, Nbins)
bins = mn + c_dx * np.arange(Nbins + 1)
return c_dx, bins
else:
return c_dx
def stadistics(self):
headers = ["group", "mean", "std dev", "min", "25%", "50%", "75%", "max", "nonzero", "nonan", "unique", "dtype"]
self.chunksize = Chunks.build_from_shape(self.shape, self.dtypes)
table = []
for group, (dtype, _) in self.dtypes.fields.items():
values = dict()
values["dtype"] = dtype
values["group"] = group
darray = self.data[group].da
if dtype == np.dtype(float) or dtype == np.dtype(int):
da_mean = da.around(darray.mean(), decimals=3)
da_std = da.around(darray.std(), decimals=3)
da_min = da.around(darray.min(), decimals=3)
da_max = da.around(darray.max(), decimals=3)
result = dask.compute([da_mean, da_std, da_min, da_max])[0]
values["mean"] = result[0] if not np.isnan(result[0]) else da.around(da.nanmean(darray), decimals=3).compute()
values["std dev"] = result[1] if not np.isnan(result[0]) else da.around(da.nanstd(darray), decimals=3).compute()
values["min"] = result[2] if not np.isnan(result[0]) else da.around(da.nanmin(darray), decimals=3).compute()
values["max"] = result[3] if not np.isnan(result[0]) else da.around(da.nanmax(darray), decimals=3).compute()
if len(self.shape[group]) == 1:
da_percentile = da.around(da.percentile(darray, [25, 50, 75]), decimals=3)
result = da_percentile.compute()
values["25%"] = result[0]
values["50%"] = result[1]
values["75%"] = result[2]
else:
values["25%"] = "-"
values["50%"] = "-"
values["75%"] = "-"
values["nonzero"] = da.count_nonzero(darray).compute()
values["nonan"] = da.count_nonzero(da.notnull(darray)).compute()
values["unique"] = "-"
else:
values["mean"] = "-"
values["std dev"] = "-"
values["min"] = "-"
values["max"] = "-"
values["25%"] = "-"
values["50%"] = "-"
values["75%"] = "-"
values["nonzero"] = "-"
values["nonan"] = da.count_nonzero(da.notnull(darray)).compute()
vunique = darray.to_dask_dataframe().fillna('').nunique().compute()
values["unique"] = vunique
row = []
for column in headers:
row.append(values[column])
table.append(row)
print("# rows {}".format(self.shape[0]))
return tabulate(table, headers)
def create_hv_dataset(ddf, stats, percentile=(1, 99)):
_idNames = ("patch", "tract", "filter")
_kdims = ("ra", "dec", "psfMag")
_flags = [c for c in ddf.columns if ddf[c].dtype == np.dtype("bool")]
kdims = []
vdims = []
for c in ddf.columns:
if c in _kdims or c in _idNames or c in _flags:
if c in ("ra", "dec", "psfMag"):
cmin, cmax = stats[c]["min"].min(), stats[c]["max"].max()
c = hv.Dimension(c, range=(cmin, cmax))
elif c in ("filter", "patch"):
cvalues = list(ddf[c].unique())
c = hv.Dimension(c, values=cvalues)
elif ddf[c].dtype.kind == "b":
c = hv.Dimension(c, values=[True, False])
kdims.append(c)
else:
if percentile is not None:
p0, p1 = percentile
if f"{p0}%" in stats.index and f"{p1}%" in stats.index:
cmin, cmax = stats[c][f"{p0}%"].min(
), stats[c][f"{p1}%"].max()
else:
print("percentiles not found in stats, computing")
darray = ddf[c].values
cmin, cmax = da.compute(
da.percentile(darray, p0)[0],
da.percentile(darray, p1)[0])
else:
cmin, cmax = stats[c]["min"].min(), stats[c]["max"].max()
c = hv.Dimension(c, range=(cmin, cmax))
vdims.append(c)
return hv.Dataset(ddf, kdims=kdims, vdims=vdims)
def gram_rbf(X, threshold=1.0):
if type(X) == torch.Tensor:
dot_products = X @ X.t()
sq_norms = dot_products.diag()
sq_distances = -2*dot_products + sq_norms[:,None] + sq_norms[None,:]
sq_median_distance = sq_distances.median()
return torch.exp(-sq_distances / (2*threshold**2 * sq_median_distance))
elif type(X) == da.Array:
dot_products = X @ X.T
sq_norms = da.diag(dot_products)
sq_distances = -2*dot_products + sq_norms[:,None] + sq_norms[None,:]
sq_median_distance = da.percentile(sq_distances.ravel(), 50)
return da.exp((-sq_distances / (2*threshold**2 * sq_median_distance)))
else:
raise ValueError
def _band_hist(band_data):
cdf = da.arange(0., 1., 1. / nwidth, chunks=nwidth)
if approximate:
# need a 1D array
flat_data = band_data.ravel()
# replace with nanpercentile in the future, if available
# dask < 0.17 returns all NaNs for this
bins = da.percentile(flat_data[da.notnull(flat_data)],
cdf * 100.)
else:
bins = dask.delayed(np.nanpercentile)(band_data, cdf * 100.)
bins = da.from_delayed(bins, shape=(nwidth,), dtype=cdf.dtype)
res = dask.delayed(np.interp)(band_data, bins, cdf)
res = da.from_delayed(res, shape=band_data.shape,
dtype=band_data.dtype)
return res
def _freedman_bw_dask(data, return_bins=True):
r"""Dask version of freedman_bin_width
Parameters
----------
data : dask array
the data
return_bins : bool (optional)
if True, then return the bin edges
Returns
-------
width : float
optimal bin width using Scott's rule
bins : ndarray
bin edges: returned if `return_bins` is True
Notes
-----
The optimal bin width is
.. math::
\Delta_b = \frac{2(q_{75} - q_{25})}{n^{1/3}}
where :math:`q_{N}` is the :math:`N` percent quartile of the data, and
:math:`n` is the number of data points.
"""
if not isinstance(data, da.Array):
raise TypeError("Expected a dask array")
if data.ndim != 1:
data = data.flatten()
n = data.size
v25, v75 = da.percentile(data, [25, 75])
dx = 2 * (v75 - v25) * n**(-1.0 / 3.0)
c_dx, mx, mn = da.compute(dx, data.max(), data.min())
if return_bins:
Nbins = max(1, np.ceil((mx - mn) / c_dx))
bins = mn + c_dx * np.arange(Nbins + 1)
return c_dx, bins
else:
return c_dx
def test_percentile():
d = da.ones((16,), chunks=(4,))
assert eq(da.percentile(d, [0, 50, 100]), [1, 1, 1])
x = np.array([0, 0, 5, 5, 5, 5, 20, 20])
d = da.from_array(x, chunks=(3,))
assert eq(da.percentile(d, [0, 50, 100]), [0, 5, 20])
assert same_keys(da.percentile(d, [0, 50, 100]),
da.percentile(d, [0, 50, 100]))
assert not same_keys(da.percentile(d, [0, 50, 100]),
da.percentile(d, [0, 50]))
x = np.array(['a', 'a', 'd', 'd', 'd', 'e'])
d = da.from_array(x, chunks=(3,))
assert eq(da.percentile(d, [0, 50, 100]), ['a', 'd', 'e'])
def _band_hist(band_data):
cdf = da.arange(0., 1., 1. / nwidth, chunks=nwidth)
if approximate:
# need a 1D array
flat_data = band_data.ravel()
# replace with nanpercentile in the future, if available
# dask < 0.17 returns all NaNs for this
bins = da.percentile(flat_data[da.notnull(flat_data)],
cdf * 100.)
else:
bins = dask.delayed(np.nanpercentile)(band_data, cdf * 100.)
bins = da.from_delayed(bins, shape=(nwidth, ), dtype=cdf.dtype)
res = dask.delayed(np.interp)(band_data, bins, cdf)
res = da.from_delayed(res,
shape=band_data.shape,
dtype=band_data.dtype)
return res
def test_percentile():
d = da.ones((16,), chunks=(4,))
assert_eq(da.percentile(d, [0, 50, 100]),
np.array([1, 1, 1], dtype=d.dtype))
x = np.array([0, 0, 5, 5, 5, 5, 20, 20])
d = da.from_array(x, chunks=(3,))
result = da.percentile(d, [0, 50, 100])
assert_eq(da.percentile(d, [0, 50, 100]),
np.array([0, 5, 20], dtype=result.dtype))
assert same_keys(da.percentile(d, [0, 50, 100]),
da.percentile(d, [0, 50, 100]))
assert not same_keys(da.percentile(d, [0, 50, 100]),
da.percentile(d, [0, 50]))
x = np.array(['a', 'a', 'd', 'd', 'd', 'e'])
d = da.from_array(x, chunks=(3,))
assert_eq(da.percentile(d, [0, 50, 100]),
np.array(['a', 'd', 'e'], dtype=x.dtype))
def test_percentile():
d = da.ones((16, ), chunks=(4, ))
assert_eq(da.percentile(d, [0, 50, 100]), np.array([1, 1, 1],
dtype=d.dtype))
x = np.array([0, 0, 5, 5, 5, 5, 20, 20])
d = da.from_array(x, chunks=(3, ))
result = da.percentile(d, [0, 50, 100])
assert_eq(da.percentile(d, [0, 50, 100]),
np.array([0, 5, 20], dtype=result.dtype))
assert same_keys(da.percentile(d, [0, 50, 100]),
da.percentile(d, [0, 50, 100]))
assert not same_keys(da.percentile(d, [0, 50, 100]),
da.percentile(d, [0, 50]))
x = np.array(['a', 'a', 'd', 'd', 'd', 'e'])
d = da.from_array(x, chunks=(3, ))
assert_eq(da.percentile(d, [0, 50, 100]),
np.array(['a', 'd', 'e'], dtype=x.dtype))
def single_channel_percentile_norm(data: da.core.Array,
min_p: float = 50.0,
max_p: float = 99.8,
**kwargs) -> da.core.Array:
# Enforce shape
if len(data.shape) > 4:
raise exceptions.InvalidShapeError(len(data.shape), 4)
# Get the norm by values
norm_by = da.percentile(data.flatten(), [min_p, min_p]).compute()
# Norm
normed = (data - norm_by[0]) / (norm_by[1] - norm_by[0])
# Clip any values outside of 0 and 1
clipped = da.clip(normed, 0, 1)
# Scale them between 0 and 255
return clipped * 255
def test_percentile(method):
d = da.ones((16,), chunks=(4,))
qs = [0, 50, 100]
assert_eq(da.percentile(d, qs, method=method),
np.array([1, 1, 1], dtype=d.dtype))
x = np.array([0, 0, 5, 5, 5, 5, 20, 20])
d = da.from_array(x, chunks=(3,))
result = da.percentile(d, qs, method=method)
assert_eq(result,
np.array([0, 5, 20], dtype=result.dtype))
assert same_keys(da.percentile(d, qs, method=method),
da.percentile(d, qs, method=method))
assert not same_keys(da.percentile(d, qs, method=method),
da.percentile(d, [0, 50], method=method))
if method != 'tdigest':
x = np.array(['a', 'a', 'd', 'd', 'd', 'e'])
d = da.from_array(x, chunks=(3,))
assert_eq(da.percentile(d, [0, 50, 100]),
np.array(['a', 'd', 'e'], dtype=x.dtype))
def from_bcolz(x, chunksize=None, categorize=True, index=None, lock=lock,
**kwargs):
""" Read dask Dataframe from bcolz.ctable
Parameters
----------
x : bcolz.ctable
Input data
chunksize : int, optional
The size of blocks to pull out from ctable. Ideally as large as can
comfortably fit in memory
categorize : bool, defaults to True
Automatically categorize all string dtypes
index : string, optional
Column to make the index
lock: bool or Lock
Lock to use when reading or False for no lock (not-thread-safe)
See Also
--------
from_array: more generic function not optimized for bcolz
"""
if lock is True:
lock = Lock()
import dask.array as da
import bcolz
if isinstance(x, (str, unicode)):
x = bcolz.ctable(rootdir=x)
bc_chunklen = max(x[name].chunklen for name in x.names)
if chunksize is None and bc_chunklen > 10000:
chunksize = bc_chunklen
categories = dict()
if categorize:
for name in x.names:
if (np.issubdtype(x.dtype[name], np.string_) or
np.issubdtype(x.dtype[name], np.unicode_) or
np.issubdtype(x.dtype[name], np.object_)):
a = da.from_array(x[name], chunks=(chunksize * len(x.names),))
categories[name] = da.unique(a)
columns = tuple(x.dtype.names)
divisions = tuple(range(0, len(x), chunksize))
divisions = divisions + (len(x) - 1,)
if x.rootdir:
token = tokenize((x.rootdir, os.path.getmtime(x.rootdir)), chunksize,
categorize, index, kwargs)
else:
token = tokenize((id(x), x.shape, x.dtype), chunksize, categorize,
index, kwargs)
new_name = 'from_bcolz-' + token
dsk = dict(((new_name, i),
(dataframe_from_ctable,
x,
(slice(i * chunksize, (i + 1) * chunksize),),
columns, categories, lock))
for i in range(0, int(ceil(len(x) / chunksize))))
meta = dataframe_from_ctable(x, slice(0, 0), columns, categories, lock)
result = DataFrame(dsk, new_name, meta, divisions)
if index:
assert index in x.names
a = da.from_array(x[index], chunks=(chunksize * len(x.names),))
q = np.linspace(0, 100, len(x) // chunksize + 2)
divisions = da.percentile(a, q).compute()
return set_partition(result, index, divisions, **kwargs)
else:
return result
def from_bcolz(x,
chunksize=None,
categorize=True,
index=None,
lock=lock,
**kwargs):
""" Read BColz CTable into a Dask Dataframe
BColz is a fast on-disk compressed column store with careful attention
given to compression. https://bcolz.readthedocs.io/en/latest/
Parameters
----------
x : bcolz.ctable
chunksize : int, optional
The size(rows) of blocks to pull out from ctable.
categorize : bool, defaults to True
Automatically categorize all string dtypes
index : string, optional
Column to make the index
lock: bool or Lock
Lock to use when reading or False for no lock (not-thread-safe)
See Also
--------
from_array: more generic function not optimized for bcolz
"""
if lock is True:
lock = Lock()
import dask.array as da
import bcolz
if isinstance(x, (str, unicode)):
x = bcolz.ctable(rootdir=x)
bc_chunklen = max(x[name].chunklen for name in x.names)
if chunksize is None and bc_chunklen > 10000:
chunksize = bc_chunklen
categories = dict()
if categorize:
for name in x.names:
if (np.issubdtype(x.dtype[name], np.string_)
or np.issubdtype(x.dtype[name], np.unicode_)
or np.issubdtype(x.dtype[name], np.object_)):
a = da.from_array(x[name], chunks=(chunksize * len(x.names), ))
categories[name] = da.unique(a)
columns = tuple(x.dtype.names)
divisions = tuple(range(0, len(x), chunksize))
divisions = divisions + (len(x) - 1, )
if x.rootdir:
token = tokenize((x.rootdir, os.path.getmtime(x.rootdir)), chunksize,
categorize, index, kwargs)
else:
token = tokenize((id(x), x.shape, x.dtype), chunksize, categorize,
index, kwargs)
new_name = 'from_bcolz-' + token
dsk = dict(((new_name, i), (dataframe_from_ctable, x,
(slice(i * chunksize, (i + 1) * chunksize), ),
columns, categories, lock))
for i in range(0, int(ceil(len(x) / chunksize))))
meta = dataframe_from_ctable(x, slice(0, 0), columns, categories, lock)
result = DataFrame(dsk, new_name, meta, divisions)
if index:
assert index in x.names
a = da.from_array(x[index], chunks=(chunksize * len(x.names), ))
q = np.linspace(0, 100, len(x) // chunksize + 2)
divisions = tuple(da.percentile(a, q).compute())
return set_partition(result, index, divisions, **kwargs)
else:
return result
def _dense_fit(self, X, random_state):
references = self.references_ * 100
quantiles = [da.percentile(col, references) for col in X.T]
self.quantiles_, = compute(da.vstack(quantiles).T)
def _dense_fit(self, X: Union[ArrayLike, DataFrameType],
random_state: int) -> Union[ArrayLike, DataFrameType]:
references = self.references_ * 100
quantiles = [da.percentile(col, references) for col in X.T]
(self.quantiles_, ) = compute(da.vstack(quantiles).T)
return None
def test_percentiles_with_empty_arrays(method):
x = da.ones(10, chunks=((5, 0, 5),))
assert_eq(da.percentile(x, [10, 50, 90], method=method), np.array([1, 1, 1], dtype=x.dtype))
def test_percentiles_with_scaler_percentile(method, q):
# Regression test to ensure da.percentile works with scalar percentiles
# See #3020
d = da.ones((16,), chunks=(4,))
assert_eq(da.percentile(d, q, method=method), np.array([1], dtype=d.dtype))
def test_percentile_tokenize():
d = da.from_array(cupy.ones((16,)), chunks=(4,))
qs = np.array([0, 50, 100])
assert same_keys(da.percentile(d, qs), da.percentile(d, qs))
def test_percentiles_with_empty_arrays():
x = da.ones(10, blockdims=((5, 0, 5),))
assert da.percentile(x, [10, 50, 90]).compute().tolist() == [1, 1, 1]