def _lu_solve_core(lu, permutation, b, trans):
m = lu.shape[0]
x = jnp.reshape(b, (m, -1))
if trans == 0:
x = x[permutation, :]
x = triangular_solve(lu,
x,
left_side=True,
lower=True,
unit_diagonal=True)
x = triangular_solve(lu, x, left_side=True, lower=False)
elif trans == 1 or trans == 2:
conj = trans == 2
x = triangular_solve(lu,
x,
left_side=True,
lower=False,
transpose_a=True,
conjugate_a=conj)
x = triangular_solve(lu,
x,
left_side=True,
lower=True,
unit_diagonal=True,
transpose_a=True,
conjugate_a=conj)
x = x[jnp.argsort(permutation), :]
else:
raise ValueError(
"'trans' value must be 0, 1, or 2, got {}".format(trans))
return lax.reshape(x, b.shape)
def csd(x, y, fs=1.0, window='hann', nperseg=None, noverlap=None, nfft=None,
detrend='constant', return_onesided=True, scaling='density',
axis=-1, average='mean'):
freqs, _, Pxy = _spectral_helper(x, y, fs, window, nperseg, noverlap, nfft,
detrend, return_onesided, scaling, axis,
mode='psd')
if y is not None:
Pxy = Pxy + 0j # Ensure complex output when x is not y
# Average over windows.
if Pxy.ndim >= 2 and Pxy.size > 0:
if Pxy.shape[-1] > 1:
if average == 'median':
bias = signal_helper._median_bias(Pxy.shape[-1]).astype(Pxy.dtype)
if jnp.iscomplexobj(Pxy):
Pxy = (jnp.median(jnp.real(Pxy), axis=-1)
+ 1j * jnp.median(jnp.imag(Pxy), axis=-1))
else:
Pxy = jnp.median(Pxy, axis=-1)
Pxy /= bias
elif average == 'mean':
Pxy = Pxy.mean(axis=-1)
else:
raise ValueError(f'average must be "median" or "mean", got {average}')
else:
Pxy = jnp.reshape(Pxy, Pxy.shape[:-1])
return freqs, Pxy
def norm(x,
ord=None,
axis: Union[None, Tuple[int, ...], int] = None,
keepdims=False):
x = _promote_arg_dtypes(jnp.asarray(x))
x_shape = jnp.shape(x)
ndim = len(x_shape)
if axis is None:
# NumPy has an undocumented behavior that admits arbitrary rank inputs if
# `ord` is None: https://github.com/numpy/numpy/issues/14215
if ord is None:
return jnp.sqrt(
jnp.sum(jnp.real(x * jnp.conj(x)), keepdims=keepdims))
axis = tuple(range(ndim))
elif isinstance(axis, tuple):
axis = tuple(canonicalize_axis(x, ndim) for x in axis)
else:
axis = (canonicalize_axis(axis, ndim), )
num_axes = len(axis)
if num_axes == 1:
if ord is None or ord == 2:
return jnp.sqrt(
jnp.sum(jnp.real(x * jnp.conj(x)),
axis=axis,
keepdims=keepdims))
elif ord == jnp.inf:
return jnp.amax(jnp.abs(x), axis=axis, keepdims=keepdims)
elif ord == -jnp.inf:
return jnp.amin(jnp.abs(x), axis=axis, keepdims=keepdims)
elif ord == 0:
return jnp.sum(x != 0,
dtype=jnp.finfo(lax.dtype(x)).dtype,
axis=axis,
keepdims=keepdims)
elif ord == 1:
# Numpy has a special case for ord == 1 as an optimization. We don't
# really need the optimization (XLA could do it for us), but the Numpy
# code has slightly different type promotion semantics, so we need a
# special case too.
return jnp.sum(jnp.abs(x), axis=axis, keepdims=keepdims)
else:
abs_x = jnp.abs(x)
ord = lax._const(abs_x, ord)
out = jnp.sum(abs_x**ord, axis=axis, keepdims=keepdims)
return jnp.power(out, 1. / ord)
elif num_axes == 2:
row_axis, col_axis = cast(Tuple[int, ...], axis)
if ord is None or ord in ('f', 'fro'):
return jnp.sqrt(
jnp.sum(jnp.real(x * jnp.conj(x)),
axis=axis,
keepdims=keepdims))
elif ord == 1:
if not keepdims and col_axis > row_axis:
col_axis -= 1
return jnp.amax(jnp.sum(jnp.abs(x),
axis=row_axis,
keepdims=keepdims),
axis=col_axis,
keepdims=keepdims)
elif ord == -1:
if not keepdims and col_axis > row_axis:
col_axis -= 1
return jnp.amin(jnp.sum(jnp.abs(x),
axis=row_axis,
keepdims=keepdims),
axis=col_axis,
keepdims=keepdims)
elif ord == jnp.inf:
if not keepdims and row_axis > col_axis:
row_axis -= 1
return jnp.amax(jnp.sum(jnp.abs(x),
axis=col_axis,
keepdims=keepdims),
axis=row_axis,
keepdims=keepdims)
elif ord == -jnp.inf:
if not keepdims and row_axis > col_axis:
row_axis -= 1
return jnp.amin(jnp.sum(jnp.abs(x),
axis=col_axis,
keepdims=keepdims),
axis=row_axis,
keepdims=keepdims)
elif ord in ('nuc', 2, -2):
x = jnp.moveaxis(x, axis, (-2, -1))
if ord == 2:
reducer = jnp.amax
elif ord == -2:
reducer = jnp.amin
else:
reducer = jnp.sum
y = reducer(svd(x, compute_uv=False), axis=-1)
if keepdims:
result_shape = list(x_shape)
result_shape[axis[0]] = 1
result_shape[axis[1]] = 1
y = jnp.reshape(y, result_shape)
return y
else:
raise ValueError("Invalid order '{}' for matrix norm.".format(ord))
else:
raise ValueError(
"Invalid axis values ({}) for jnp.linalg.norm.".format(axis))
