# Copyright (c) 2021. Lucas G. S. Jeub
#
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
from pathlib import Path
import numpy as np
from dask.distributed import secede, rejoin
from local2global.utils import relative_scale, relative_orthogonal_transform
from local2global.patch import Patch
from local2global_embedding.run.scripts.utils import num_patches, num_nodes, get_dim
[docs]
def temporal_align_errors(patches, output_file, scale=True):
output_file = Path(output_file)
if not output_file.is_file():
secede()
n_nodes = num_nodes(patches).compute()
n_patches = num_patches(patches).compute()
patch = patches[0].compute()
rejoin()
dim = patch.shape[1]
reference = np.zeros((n_nodes, dim))
counts = np.zeros((n_nodes), dtype=np.int32)
counts[patch.nodes] = 1
reference[patch.nodes] = patch.coordinates
workfile = output_file.with_suffix('.tmp.npy')
try:
errors = np.lib.format.open_memmap(workfile, mode='w+', dtype=float, shape=(n_nodes, n_patches))
errors[:, :] = np.nan
errors[patch.nodes, 0] = 0.0
for pi in range(1, n_patches):
secede()
patch = patches[pi].compute()
rejoin()
valid_nodes = patch.nodes[counts[patch.nodes] > 0]
ref = reference[valid_nodes]
coords = patch.get_coordinates(valid_nodes)
if scale:
scale_factor = relative_scale(ref, coords)
patch.coordinates *= scale_factor
rot = relative_orthogonal_transform(ref, coords)
patch.coordinates = patch.coordinates @ rot.T
patch.coordinates -= np.nanmean(coords, axis=0, keepdims=True)
patch.coordinates += np.nanmean(ref, axis=0, keepdims=True)
errors[valid_nodes, pi] = np.linalg.norm(ref - patch.get_coordinates(valid_nodes), axis=1)
reference[patch.nodes] *= counts[patch.nodes, None]
counts[patch.nodes] += 1
reference[patch.nodes] += patch.coordinates
reference[patch.nodes] /= counts[patch.nodes, None]
errors.flush()
workfile.replace(output_file)
finally:
workfile.unlink(missing_ok=True)
return output_file