def test_apply_cut(s):
cut = Cut((0, 1), (2, ))
cut_s = s.apply_cut(cut)
assert s.network == cut_s.network
assert s.state == cut_s.state
assert s.node_indices == cut_s.node_indices
assert np.array_equal(cut_s.tpm, s.tpm)
assert np.array_equal(cut_s.cm, cut.apply_cut(s.cm))
def test_apply_cut(s):
cut = Cut((0, 1), (2,))
cut_s = s.apply_cut(cut)
assert s.network == cut_s.network
assert s.state == cut_s.state
assert s.node_indices == cut_s.node_indices
assert np.array_equal(cut_s.tpm, s.tpm)
assert np.array_equal(cut_s.cm, cut.apply_cut(s.cm))
def test_validate_cut_nodes_equal_subsystem_nodes(s):
assert s.node_indices == (0, 1, 2)
cut = Cut((0, ), (1, 2)) # A-ok
Subsystem(s.network, s.state, s.node_indices, cut=cut)
cut = Cut((0, ), (1, )) # missing node 2 in cut
with pytest.raises(ValueError):
Subsystem(s.network, s.state, s.node_indices, cut=cut)
cut = Cut((0, ), (1, 2)) # missing node 2 in subsystem
with pytest.raises(ValueError):
Subsystem(s.network, s.state, (0, 1), cut=cut)
def test_find_cut_matrix(s, big_subsys_0_thru_3):
cut = Cut((0, ), (1, 2))
cut_s = Subsystem(s.node_indices,
s.network,
cut=cut,
mice_cache=s._mice_cache)
answer_s = np.array([[0, 1, 1], [0, 0, 0], [0, 0, 0]])
cut = Cut((0, 1), (2, 3))
cut_big = Subsystem(big_subsys_0_thru_3.node_indices,
big_subsys_0_thru_3.network,
cut=cut,
mice_cache=big_subsys_0_thru_3._mice_cache)
answer_big = np.array([[0, 0, 1, 1], [0, 0, 1, 1], [0, 0, 0, 0],
[0, 0, 0, 0]])
assert np.array_equal(cut_s.cut_matrix, answer_s)
assert np.array_equal(cut_big.cut_matrix, answer_big)
def test_find_mip_parallel_standard_example(s, flushcache, restore_fs_cache):
flushcache()
unpartitioned_constellation = constellation(s)
bipartitions = utils.directed_bipartition(s.node_indices)[1:-1]
cuts = [Cut(bipartition[0], bipartition[1])
for bipartition in bipartitions]
min_mip = _null_bigmip(s)
min_mip.phi = float('inf')
mip = _find_mip_parallel(s, cuts, unpartitioned_constellation, min_mip)
check_mip(mip, standard_answer)
def test_find_mip_sequential_micro(micro_s, flushcache, restore_fs_cache):
flushcache()
unpartitioned_constellation = constellation(micro_s)
bipartitions = utils.directed_bipartition(micro_s.node_indices)[1:-1]
cuts = [Cut(bipartition[0], bipartition[1])
for bipartition in bipartitions]
min_mip = _null_bigmip(micro_s)
min_mip.phi = float('inf')
mip = _find_mip_sequential(micro_s, cuts, unpartitioned_constellation,
min_mip)
check_mip(mip, micro_answer)
def test_find_cut_matrix(s, big_subsys_0_thru_3):
cut = Cut((0, ), (1, 2))
cut_s = Subsystem(s.network, s.state, s.node_indices, cut=cut)
answer_s = np.array([[0, 1, 1], [0, 0, 0], [0, 0, 0]])
assert np.array_equal(cut_s.cut_matrix, answer_s)
cut = Cut((0, 1), (2, 3))
cut_big = Subsystem(big_subsys_0_thru_3.network,
big_subsys_0_thru_3.state,
big_subsys_0_thru_3.node_indices,
cut=cut)
answer_big = np.array([[0, 0, 1, 1], [0, 0, 1, 1], [0, 0, 0, 0],
[0, 0, 0, 0]])
assert np.array_equal(cut_big.cut_matrix, answer_big)
null_cut_matrix = np.array([
[0, 0, 0],
[0, 0, 0],
[0, 0, 0],
])
assert np.array_equal(s.cut_matrix, null_cut_matrix)
def test_find_mip_parallel_standard_example(s, flushcache, restore_fs_cache):
flushcache()
initial = (config.PARALLEL_CUT_EVALUATION, config.NUMBER_OF_CORES)
config.PARALLEL_CUT_EVALUATION, config.NUMBER_OF_CORES = True, -2
unpartitioned_constellation = constellation(s)
bipartitions = utils.directed_bipartition(s.node_indices)[1:-1]
cuts = [Cut(bipartition[0], bipartition[1])
for bipartition in bipartitions]
min_mip = _null_bigmip(s)
min_mip.phi = float('inf')
mip = _find_mip_parallel(s, cuts, unpartitioned_constellation, min_mip)
check_mip(mip, standard_answer)
config.PARALLEL_CUT_EVALUATION, config.NUMBER_OF_CORES = initial
def test_find_mip_sequential_noised_example(s_noised, flushcache,
restore_fs_cache):
flushcache()
initial = config.PARALLEL_CUT_EVALUATION
config.PARALLEL_CUT_EVALUATION = False
unpartitioned_constellation = constellation(s_noised)
bipartitions = utils.directed_bipartition(s_noised.node_indices)[1:-1]
cuts = [Cut(bipartition[0], bipartition[1])
for bipartition in bipartitions]
min_mip = _null_bigmip(s_noised)
min_mip.phi = float('inf')
mip = _find_mip_sequential(s_noised, cuts, unpartitioned_constellation, min_mip)
check_mip(mip, noised_answer)
config.PARALLEL_CUT_EVALUATION = initial
def test_find_mip_parallel_micro(micro_s, flushcache, restore_fs_cache):
flushcache()
initial = config.PARALLEL_CUT_EVALUATION
config.PARALLEL_CUT_EVALUATION = True
unpartitioned_constellation = constellation(micro_s)
bipartitions = utils.directed_bipartition(micro_s.node_indices)[1:-1]
cuts = [Cut(bipartition[0], bipartition[1])
for bipartition in bipartitions]
min_mip = _null_bigmip(micro_s)
min_mip.phi = float('inf')
mip = _find_mip_parallel(micro_s, cuts, unpartitioned_constellation,
min_mip)
check_mip(mip, micro_answer)
config.PARALLEL_CUT_EVALUATION = initial
def big_mip_bipartitions(nodes):
"""pyphi.compute.big_mip_bipartitions()
Return all |big_phi| cuts for the given nodes.
This value changes based on `config.CUT_ONE_APPROXIMATION`.
Args:
nodes tuple(int): The node indices to partition.
Returns:
list(Cut): All unidirectional partitions.
"""
if config.CUT_ONE_APPROXIMATION:
bipartitions = utils.directed_bipartition_of_one(nodes)
else:
# Skip the first and last (trivial, null cut) bipartitions
bipartitions = utils.directed_bipartition(nodes)[1:-1]
return [
Cut(bipartition[0], bipartition[1]) for bipartition in bipartitions
]
import pytest
from itertools import chain
from pyphi import Subsystem
from pyphi.models import Mice, Cut
from pyphi.utils import phi_eq
import example_networks
# Expected results {{{
# ====================
s = example_networks.s()
directions = ('past', 'future')
cuts = (None, Cut((1, 2), (0, )))
subsystem = {
cut: Subsystem(s.node_indices, s.network, cut=cut)
for cut in cuts
}
expected_purview_indices = {
cuts[0]: {
'past': {
(1, ): (2, ),
(2, ): (0, 1),
(0, 1): (1, 2),
(0, 1, 2): (0, 1, 2)
},
'future': {
(1, ): (0, ),
[2],
np.array([0.5, 0.5]).reshape(1, 1, 2, order="F")
), (
'cause_repertoire',
standard_subsystem,
[1],
[],
np.array([1])
),
# }}}
# Full network, with cut {{{
# --------------------------
(
'cause_repertoire',
Subsystem(standard, standard_subsystem.state, full,
cut=Cut((2,), (0, 1))),
[0],
[1],
np.array([1 / 3, 2 / 3]).reshape(1, 2, 1, order="F")
),
# }}}
# Subset, with cut {{{
# --------------------
(
'cause_repertoire',
Subsystem(standard, standard_subsystem.state, (1, 2),
cut=Cut((1,), (2,))),
[2],
[1, 2],
np.array([0.25, 0.25, 0.25, 0.25]).reshape(1, 2, 2, order="F")
), (
def test_cut_mechanisms(s):
assert list(s.cut_mechanisms) == []
assert list(s.apply_cut(Cut((0, 1), (2,))).cut_mechanisms) == [
(0, 2), (1, 2), (0, 1, 2)]
def test_cut_mechanisms(s):
assert s.cut_mechanisms == ()
assert s.apply_cut(Cut(
(0, 1), (2, ))).cut_mechanisms == ((0, 2), (1, 2), (0, 1, 2))
def test_is_cut(s):
assert s.is_cut is False
s = Subsystem(s.network, s.state, s.node_indices, cut=Cut((0, ), (1, 2)))
assert s.is_cut is True
np.array([0.5, 0.5]).reshape(2, 1, 1, order="F")),
('cause_repertoire', standard_subsystem, [0, 1], [0, 2],
np.array([0.5, 0.5, 0.0, 0.0]).reshape(2, 1, 2, order="F")),
('cause_repertoire', standard_subsystem, [1], [2],
np.array([1.0, 0.0]).reshape(1, 1, 2, order="F")),
('cause_repertoire', standard_subsystem, [], [2],
np.array([0.5, 0.5]).reshape(1, 1, 2, order="F")),
('cause_repertoire', standard_subsystem, [1], [], np.array([1])),
# }}}
# Full network, with cut {{{
# --------------------------
('cause_repertoire',
Subsystem(standard,
standard_subsystem.state,
full,
cut=Cut((2, ), (0, 1))), [0], [1],
np.array([1 / 3, 2 / 3]).reshape(1, 2, 1, order="F")),
# }}}
# Subset, with cut {{{
# --------------------
('cause_repertoire',
Subsystem(standard,
standard_subsystem.state, (1, 2),
cut=Cut((1, ), (2, ))), [2], [1, 2],
np.array([0.25, 0.25, 0.25, 0.25]).reshape(1, 2, 2, order="F")),
('cause_repertoire',
Subsystem(standard,
standard_subsystem.state, (1, 2),
cut=Cut((1, ),
(2, ))), [2], [2], np.array([0.5,
0.5]).reshape(1,
('cause_repertoire', Subsystem(full, standard, cut=None), [0], [0],
np.array([0.5, 0.5]).reshape(2, 1, 1, order="F")),
('cause_repertoire', Subsystem(full, standard, cut=None), [0], [0],
np.array([0.5, 0.5]).reshape(2, 1, 1, order="F")),
('cause_repertoire', Subsystem(full, standard, cut=None), [0, 1], [0, 2],
np.array([0.5, 0.5, 0.0, 0.0]).reshape(2, 1, 2, order="F")),
('cause_repertoire', Subsystem(full, standard, cut=None), [1], [2],
np.array([1.0, 0.0]).reshape(1, 1, 2, order="F")),
('cause_repertoire', Subsystem(full, standard, cut=None), [], [2],
np.array([0.5, 0.5]).reshape(1, 1, 2, order="F")),
('cause_repertoire', Subsystem(full, standard,
cut=None), [1], [], np.array([1])),
# }}}
# Full network, with cut {{{
# --------------------------
('cause_repertoire', Subsystem(full, standard, cut=Cut(
(2, ), (0, 1))), [0], [1], np.array([1 / 3,
2 / 3]).reshape(1,
2,
1,
order="F")),
# }}}
# Subset, with cut {{{
# --------------------
('cause_repertoire', Subsystem((1, 2), standard, cut=Cut(
(1, ), (2, ))), [2], [1, 2], np.array([0.25, 0.25, 0.25,
0.25]).reshape(1,
2,
2,
order="F")),
('cause_repertoire', Subsystem((1, 2), standard, cut=Cut(
(1, ), (2, ))), [2], [2], np.array([0.5, 0.5]).reshape(1,