1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
|
"""This module contains various utility functions and classes for the MOArchiving package."""
import warnings as _warnings
try:
from sortedcontainers import SortedKeyList
except ImportError:
_warnings.warn('`sortedcontainers` module not installed, moarchiving for 3 and 4 objectives will not work')
SortedKeyList = list
class DLNode:
""" A class to represent a node in a doubly linked list. """
def __init__(self, x=None, info=None):
""" Initialize a node with the given x-coordinate and info. """
self.x = x if x else [None, None, None, None]
self.closest = [None, None] # closest in x coordinate, closest in y coordinate
self.cnext = [None, None] # current next
self.next = [None, None, None, None]
self.prev = [None, None, None, None]
self.ndomr = 0 # number of dominators
self.info = info
def copy(self):
""" copy the node """
new_node = DLNode()
for var in self.__dict__:
if isinstance(getattr(self, var), list):
setattr(new_node, var, getattr(self, var).copy())
else:
setattr(new_node, var, getattr(self, var))
return new_node
class ArchiveSortedList(SortedKeyList):
""" A class to represent a sorted list of nodes, together with additional methods that
follow the definition in the paper."""
def __init__(self, iterable=None, key=lambda node: node.x[1]):
""" Initialize the sorted list with the given iterable and key function. """
if SortedKeyList is list:
raise ImportError("`MySortedList `requires `sortedcontainers` to be installed")
super().__init__(iterable=iterable, key=key)
def __str__(self):
""" Return a string representation of the sorted list. """
return str([node.x for node in self])
def head_y(self):
""" Return the point q from the list, with the smallest q_y """
return self[0]
def head_x(self):
""" Return the point q from the list, with the smallest q_x """
return self[-1]
def next_y(self, s):
""" Return the point q from the list, with the smallest q_y > s_y, for a given point s
from the list """
return self[self.index(s) + 1]
def next_x(self, s):
""" Return the point q from the list, with the smallest q_x > s_x, for a given point s
from the list """
return self[self.index(s) - 1]
def outer_delimiter_y(self, p):
""" Return the point q from the list, with the smallest q_y > p_y, such that q_x < p_x """
i = self.bisect_left(p)
while i < len(self) and self[i].x[0] >= p.x[0]:
i += 1
return self[i]
def outer_delimiter_x(self, p):
""" Return the point q from the list, with the smallest q_x > p_x, such that q_y < p_y """
i = self.bisect_left(p) - 1
while i > 0 and self[i].x[1] >= p.x[1]:
i -= 1
return self[i]
def remove_dominated_y(self, p, s):
""" For s = outer_delimiter_x(p), remove all points q, such that p* <= q* from the list,
and return them sorted by ascending order of q_y """
e = self.next_y(s)
points_to_remove = []
while p.x[0] <= e.x[0]:
points_to_remove.append(e)
e = self.next_y(e)
for q in points_to_remove:
self.remove(q)
return points_to_remove
def remove_dominated_x(self, p, s):
""" For s = outer_delimiter_y(p), remove all points q, such that p* <= q* from the list,
and return them sorted by ascending order of q_x """
e = self.next_x(s)
points_to_remove = []
while p.x[1] <= e.x[1]:
points_to_remove.append(e)
e = self.next_x(e)
for q in points_to_remove:
self.remove(q)
return points_to_remove
def add_y(self, p, s):
""" Insert point p into the list, if s_y < p_y < next_y(s)_y or p_y < head_y_y """
if len(self) == 0:
self.add(p)
elif s.x[1] < p.x[1] < self.next_y(s).x[1]:
self.add(p)
elif p.x[1] < self.head_y().x[1] and s is None:
self.add(p)
def add_x(self, p, s):
""" Insert point p into the list, if s_x < p_x < next_x(s)_x or p_x < head_x_x """
if len(self) == 0:
self.add(p)
elif s.x[0] < p.x[0] < self.next_x(s).x[0]:
self.add(p)
elif p.x[0] < self.head_x().x[0] and s is None:
self.add(p)
def my_lexsort(keys):
""" Sort an array of keys in lexicographic order and return the indices.
Equivalent to np.lexsort """
idk_key_tuple = list(enumerate([list(x)[::-1] for x in zip(*keys)]))
idk_key_tuple.sort(key=lambda x: x[1])
return [x[0] for x in idk_key_tuple]
# --------------- Auxiliary Functions ---------------------
def lexicographic_less(a, b):
""" Returns True if a is lexicographically less than b, False otherwise """
return a[2] < b[2] or (a[2] == b[2] and (a[1] < b[1] or (a[1] == b[1] and a[0] <= b[0])))
def init_sentinels_new(list_nodes, ref, dim):
""" Initialize the sentinel nodes for the list of nodes given
the reference point and the dimensionality """
s1, s2, s3 = list_nodes[0], list_nodes[1], list_nodes[2]
# Initialize s1 node
s1.x = [float('-inf'), ref[1], float('-inf'), float('-inf')]
s1.closest = [s2, s1]
s1.next = [None, None, s2, s2]
s1.cnext = [None, None]
s1.prev = [None, None, s3, s3]
s1.ndomr = 0
# Initialize s2 node
s2.x = [ref[0], float('-inf'), float('-inf'), float('-inf')]
s2.closest = [s2, s1]
s2.next = [None, None, s3, s3]
s2.cnext = [None, None]
s2.prev = [None, None, s1, s1]
s2.ndomr = 0
# Initialize s3 node
s3.x = [float('-inf'), float('-inf'), ref[2], ref[3] if dim == 4 else float('-inf')]
s3.closest = [s2, s1]
s3.next = [None, None, s1, None]
s3.cnext = [None, None]
s3.prev = [None, None, s2, s2]
s3.ndomr = 0
return s1
def add_to_z(new):
""" Add a new node to the list sorted by z """
new.next[2] = new.prev[2].next[2]
new.next[2].prev[2] = new
new.prev[2].next[2] = new
def remove_from_z(old, archive_dim):
""" Remove a node from the list sorted by z """
di = archive_dim - 1
old.prev[di].next[di] = old.next[di]
old.next[di].prev[di] = old.prev[di]
def setup_z_and_closest(head, new):
""" Sets up the closest[0] and closest[1] pointers for the new node """
closest1 = head
closest0 = head.next[2]
q = head.next[2].next[2]
newx = new.x
while q and lexicographic_less(q.x, newx):
if q.x[0] <= newx[0] and q.x[1] <= newx[1]:
new.ndomr += 1
elif q.x[1] < newx[1] and (
q.x[0] < closest0.x[0] or (q.x[0] == closest0.x[0] and q.x[1] < closest0.x[1])):
closest0 = q
elif q.x[0] < newx[0] and (
q.x[1] < closest1.x[1] or (q.x[1] == closest1.x[1] and q.x[0] < closest1.x[0])):
closest1 = q
q = q.next[2]
new.closest[0] = new.cnext[0] = closest0
new.closest[1] = new.cnext[1] = closest1
new.prev[2] = q.prev[2] if q else None
new.next[2] = q
def update_links(head, new, p):
stop = head.prev[2]
ndom = 0
all_delimiters_visited = False
while p != stop and not all_delimiters_visited:
if p.x[0] <= new.x[0] and p.x[1] <= new.x[1] and (p.x[0] < new.x[0] or p.x[1] < new.x[1]):
all_delimiters_visited = True
else:
if new.x[0] <= p.x[0]:
if new.x[1] <= p.x[1]:
p.ndomr += 1
ndom += 1
remove_from_z(p, 3)
elif new.x[0] < p.x[0] and (new.x[1] < p.closest[1].x[1] or (
new.x[1] == p.closest[1].x[1] and (new.x[0] < p.closest[1].x[0] or (
new.x[0] == p.closest[1].x[0] and new.x[2] < p.closest[1].x[2])))):
p.closest[1] = new
elif new.x[1] < p.x[1] and (new.x[0] < p.closest[0].x[0] or (
new.x[0] == p.closest[0].x[0] and (new.x[1] < p.closest[0].x[1] or (
new.x[1] == p.closest[0].x[1] and new.x[2] < p.closest[0].x[2])))):
p.closest[0] = new
p = p.next[2]
return ndom
def restart_list_y(head):
""" Resets the cnext pointers for the y-dimension."""
head.next[2].cnext[1] = head
head.cnext[0] = head.next[2]
def compute_area_simple(p, di, s, u, Fc):
""" Computes the area as described in the paper """
dj = 1 - di
area = Fc(0)
q = s
area += (Fc(q.x[dj]) - Fc(p[dj])) * (Fc(u.x[di]) - Fc(p[di]))
while p[dj] < u.x[dj]:
q = u
u = u.cnext[di]
area += (Fc(q.x[dj]) - Fc(p[dj])) * (Fc(u.x[di]) - Fc(q.x[di]))
return area
def restart_base_setup_z_and_closest(head, new):
# Sets up closest[0] and closest[1] for the new node
p = head.next[2].next[2]
closest1 = head
closest0 = head.next[2]
newx = new.x
restart_list_y(head)
while p and lexicographic_less(p.x, newx):
p.cnext[0] = p.closest[0]
p.cnext[1] = p.closest[1]
p.cnext[0].cnext[1] = p
p.cnext[1].cnext[0] = p
if p.x[0] <= newx[0] and p.x[1] <= newx[1]:
new.ndomr += 1
elif p.x[1] < newx[1] and (
p.x[0] < closest0.x[0] or (p.x[0] == closest0.x[0] and p.x[1] < closest0.x[1])):
closest0 = p
elif p.x[0] < newx[0] and (
p.x[1] < closest1.x[1] or (p.x[1] == closest1.x[1] and p.x[0] < closest1.x[0])):
closest1 = p
p = p.next[2]
new.closest[0] = closest0
new.closest[1] = closest1
new.prev[2] = p.prev[2] if p else None
new.next[2] = p
def one_contribution_3_obj(head, new, Fc):
""" Computes the contribution of adding a new point to the archive in three dimensions """
restart_base_setup_z_and_closest(head, new)
if new.ndomr > 0:
return 0
new.cnext[0] = new.closest[0]
new.cnext[1] = new.closest[1]
area = compute_area_simple(new.x, 1, new.cnext[0], new.cnext[0].cnext[1], Fc)
p = new.next[2]
lastz = Fc(new.x[2])
volume = Fc(0)
while p and (p.x[0] > new.x[0] or p.x[1] > new.x[1]):
volume += area * (Fc(p.x[2]) - lastz)
p.cnext[0] = p.closest[0]
p.cnext[1] = p.closest[1]
if p.x[0] >= new.x[0] and p.x[1] >= new.x[1]:
area -= compute_area_simple(p.x, 1, p.cnext[0], p.cnext[0].cnext[1], Fc)
p.cnext[1].cnext[0] = p
p.cnext[0].cnext[1] = p
elif p.x[0] >= new.x[0]:
if p.x[0] <= new.cnext[0].x[0]:
x = [p.x[0], new.x[1], p.x[2]]
area -= compute_area_simple(x, 1, new.cnext[0], new.cnext[0].cnext[1], Fc)
p.cnext[0] = new.cnext[0]
p.cnext[1].cnext[0] = p
new.cnext[0] = p
else:
if p.x[1] <= new.cnext[1].x[1]:
x = [new.x[0], p.x[1], p.x[2]]
area -= compute_area_simple(x, 0, new.cnext[1], new.cnext[1].cnext[0], Fc)
p.cnext[1] = new.cnext[1]
p.cnext[0].cnext[1] = p
new.cnext[1] = p
lastz = p.x[2]
p = p.next[2]
if p:
volume += area * (Fc(p.x[2]) - Fc(lastz))
return volume
def setup_cdllist(n_obj, points, ref, infos):
""" Set up a circular doubly linked list from the given data and reference point """
points = [p for p in points if strictly_dominates(p, ref, n_obj)]
n = len(points)
head = [DLNode(info=info) for info in ["s1", "s2", "s3"] + [None] * n]
# init_sentinels_new accepts a list at the beginning, therefore we use head[0:3]
init_sentinels_new(head[0:3], ref, n_obj)
di = n_obj - 1 # Dimension index for sorting (z-axis in 3D)
if n > 0:
# Convert data to a structured format suitable for sorting and linking
if n_obj == 3:
# Using lexsort to sort by z, y, x in ascending order
sorted_indices = my_lexsort(([p[0] for p in points], [p[1] for p in points],
[p[2] for p in points]))
elif n_obj == 4:
# Using lexsort to sort by w, z, y, x in ascending order
sorted_indices = my_lexsort(([p[0] for p in points], [p[1] for p in points],
[p[2] for p in points], [p[3] for p in points]))
else:
raise ValueError("Only 3D and 4D points are supported")
# Create nodes from sorted points
for i, index in enumerate(sorted_indices):
head[i + 3].x = points[index]
head[i + 3].info = infos[index]
if n_obj == 3:
# Add 0.0 for 3d points so that it matches the original C code
head[i + 3].x.append(0.0)
# Link nodes
s = head[0].next[di]
s.next[di] = head[3]
head[3].prev[di] = s
for i in range(3, n + 2):
head[i].next[di] = head[i + 1] if i + 1 < len(head) else head[0]
head[i + 1].prev[di] = head[i]
s = head[0].prev[di]
s.prev[di] = head[n + 2]
head[n + 2].next[di] = s
return head[0]
def weakly_dominates(a, b, n_obj):
""" Return True if a weakly dominates b, False otherwise
>>> weakly_dominates([1, 2, 3], [2, 3, 3], n_obj=3)
True
>>> weakly_dominates([1, 2, 3], [2, 2, 2], n_obj=3)
False
>>> weakly_dominates([1, 2, 3], [1, 2, 3], n_obj=3)
True
"""
return all(a[i] <= b[i] for i in range(n_obj))
def strictly_dominates(a, b, n_obj):
""" Return True if a strictly dominates b, False otherwise
>>> strictly_dominates([1, 2, 3], [2, 3, 3], n_obj=3)
True
>>> strictly_dominates([1, 2, 3], [2, 2, 2], n_obj=3)
False
>>> strictly_dominates([1, 2, 3], [1, 2, 3], n_obj=3)
False
"""
return (all(a[i] <= b[i] for i in range(n_obj)) and
any(a[i] < b[i] for i in range(n_obj)))
def hv3dplus(head, Fc):
""" Computes the hypervolume indicator in d=3 in linear time """
p = head
area = Fc(0)
volume = Fc(0)
restart_list_y(head)
p = p.next[2].next[2]
stop = head.prev[2]
while p != stop:
if p.ndomr < 1:
p.cnext[0] = p.closest[0]
p.cnext[1] = p.closest[1]
area += compute_area_simple(p.x, 1, p.cnext[0], p.cnext[0].cnext[1], Fc)
p.cnext[0].cnext[1] = p
p.cnext[1].cnext[0] = p
else:
remove_from_z(p, 3)
volume += area * (Fc(p.next[2].x[2]) - Fc(p.x[2]))
p = p.next[2]
return volume
def hv4dplusR(head, Fc):
""" Compute the hypervolume indicator in d=4 by iteratively
computing the hypervolume indicator in d=3 (using hv3d+) """
hv = Fc(0)
stop = head.prev[3]
new = head.next[3].next[3]
while new != stop:
setup_z_and_closest(head, new) # Compute cx and cy of 'new' and determine next and prev in z
add_to_z(new) # Add 'new' to list sorted by z
update_links(head, new, new.next[2]) # Update cx and cy of the points above 'new' in z
# and remove dominated points
volume = hv3dplus(head, Fc) # Compute hv indicator in d=3 in linear time
height = Fc(new.next[3].x[3]) - Fc(new.x[3])
hv += volume * height # Update hypervolume in d=4
new = new.next[3]
return hv
def hv4dplusU(head, Fc):
""" Compute the hypervolume indicator in d=4 by iteratively
computing the one contribution problem in d=3.
"""
volume = Fc(0)
hv = Fc(0)
last = head.prev[3]
new = head.next[3].next[3]
while new != last:
volume += one_contribution_3_obj(head, new, Fc)
add_to_z(new)
update_links(head, new, new.next[2])
height = Fc(new.next[3].x[3]) - Fc(new.x[3])
hv += volume * height
new = new.next[3]
return hv
|