Source code for cr.sparse._src.fom.util
# Copyright 2021 CR-Suite Development Team
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from jax import jit
import jax.numpy as jnp
import cr.nimble as cnb
import cr.sparse as crs
from cr.nimble import AH_v
[docs]def matrix_affine_func(A=None, b=None):
"""Returns an affine function for a matrix A and vector b
"""
if A is not None:
A = jnp.asarray(A)
A = cnb.promote_arg_dtypes(A)
if b is not None:
b = jnp.asarray(b)
b = cnb.promote_arg_dtypes(b)
@jit
def identity(x):
# both A and b are unspecified.
return x
@jit
def translate(x):
# only b is specified.
return x + b
@jit
def similar(x):
# only A is specified
return A @ x
@jit
def affine(x):
# both A and b are specified
return A @ x + b
ax_plus_b = identity
if A is None:
# We assume that A is identity
ax_plus_b = translate
elif b is None:
# we compute y = A @ x
ax_plus_b = similar
else:
# we compute A @ x + b
ax_plus_b = affine
@partial(jit, static_argnums=(2,))
def operator(x, mode=0):
if mode == 0:
return A @ x
if mode == 1:
return AH_v(A, x)
if mode == 2:
return ax_plus_b(x)
return operator