def _check_intersect(s, h1, h2):
# When two houses intersect,
# some numbers cannot at the same time be outside of the intersection
# Thus they must be inside the intersection
changed = 0
log.append("In House {0} & {1}:", h1, h2)
h1u = util.difference(h1, h2) # h1 unique tiles
h2u = util.difference(h2, h1) # h2 unique tiles
h1un = s.get_numbers(h1u) # h1 unique numbers
h2un = s.get_numbers(h2u) # h2 unique numbers
h1i = Sudoku.other(h1un) # h1 intersect-only numbers
h2i = Sudoku.other(h2un) # h2 intersect-only numbers
hi = util.union(h1i, h2i) # intersect-only numbers
if len(hi) > 0:
@log.rollback
def rem():
log.indent()
log.append("Intersection contains {0}", hi)
log.indent()
result = max(section_remove(s, h2u, h1i), section_remove(s, h1u, h2i))
log.dedent(2)
return result
changed = rem()
return changed
def BOTTOM_INTRO(goal, assumptions, size):
"""
Proofs of the form
prove <#> via bottom-intro:
prove <[prop]> via [rule]: ...
prove <~[prop]> via [rule]: ...
"""
# try instantiating [prop] with all assumed props and subprops
all_props = union({prop, *prop.subprops} for prop in assumptions)
for prop in all_props:
if prop.kind == PropKind.NOT:
unwrapped = prop.contained
for prop_proof_size, unwrapped_proof_size in share(size - 2, 2):
prop_proof = find_proof(prop, assumptions, prop_proof_size)
unwrapped_proof = find_proof(unwrapped, assumptions,
unwrapped_proof_size)
if prop_proof is not None and unwrapped_proof is not None:
return [unwrapped_proof, prop_proof]
else:
negated = Prop(PropKind.NOT, prop)
for prop_proof_size, negated_proof_size in share(size - 2, 2):
prop_proof = find_proof(prop, assumptions, prop_proof_size)
negated_proof = find_proof(negated, assumptions,
negated_proof_size)
if prop_proof is not None and negated_proof is not None:
return [prop_proof, negated_proof]
def sentas_to_deltas(self, sentas: Iterable[SentA]) \
-> Iterable[Delta]:
# Find max incoming activation and sum of incoming activations for
# each node.
maxin_d: Dict[Node, float] = defaultdict(float) # positive influences
possumin_d: Dict[Node, float] = defaultdict(float)
minin_d: Dict[Node, float] = defaultdict(float) # negative influences
negsumin_d: Dict[Node, float] = defaultdict(float)
for senta in sentas:
if senta.a >= 0:
maxin_d[senta.to_node] = max(maxin_d[senta.to_node], senta.a)
possumin_d[senta.to_node] += senta.a
elif senta.a < 0:
minin_d[senta.to_node] = min(minin_d[senta.to_node], senta.a)
negsumin_d[senta.to_node] += senta.a
# Make Deltas from the Tyrrell averages of the SentA's
multiplier = 1.0 - self.alpha
for node in union(maxin_d.keys(), minin_d.keys()):
amt = multiplier * (
((maxin_d.get(node, 0.0) +
self.tyrrell_alpha * possumin_d.get(node, 0.0)) /
(1 + self.tyrrell_alpha)) +
((minin_d.get(node, 0.0) +
self.tyrrell_beta * negsumin_d.get(node, 0.0)) /
(1 + self.tyrrell_beta)))
if abs(amt) >= epsilon:
yield Delta(node, amt)
def propagate_once(self, g, old_d):
# Decay.
new_d: Dict[Node, float] = defaultdict(
float, ((node, self.clip_a(g, node, a * self.alpha))
for node, a in old_d.items()))
# TODO Remove nodes with a < epsilon
# Find max incoming activation and sum of incoming activations for
# each node.
maxin_d: Dict[Node, float] = defaultdict(float) # positive influences
possumin_d: Dict[Node, float] = defaultdict(float)
minin_d: Dict[Node, float] = defaultdict(float) # negative influences
negsumin_d: Dict[Node, float] = defaultdict(float)
for delta in self.make_deltas(g, old_d):
amt = (
delta.amt *
(1.0 +
self.positive_feedback_rate * old_d.get(delta.nodeid, 0.0)) *
(1.0 - self.alpha))
'''
if delta.nodeid == Before(7) and delta.amt < 0.0:
print()
print('DE', delta, ' ', amt)
print()
'''
if amt >= epsilon:
maxin_d[delta.nodeid] = max(maxin_d[delta.nodeid], amt)
possumin_d[delta.nodeid] += amt
elif amt <= -epsilon:
minin_d[delta.nodeid] = min(minin_d[delta.nodeid], amt)
negsumin_d[delta.nodeid] += amt
# Apply the Tyrrell averages of the deltas
for node in union(maxin_d.keys(), minin_d.keys()):
#print('PR', node, maxin_d.get(node, 0.0), possumin_d.get(node, 0.0), minin_d.get(node, 0.0), negsumin_d.get(node, 0.0))
'''
print('PR1', node, minin_d.get(node, 0.0), negsumin_d.get(node, 0.0), (
(minin_d.get(node, 0.0)
+ self.tyrrell_beta * negsumin_d.get(node, 0.0))
/
(1 + self.tyrrell_beta)
))
'''
new_a = new_d[node] + (
(maxin_d.get(node, 0.0) +
self.tyrrell_alpha * possumin_d.get(node, 0.0)) /
(1 + self.tyrrell_alpha)) + (
(minin_d.get(node, 0.0) +
self.tyrrell_beta * negsumin_d.get(node, 0.0)) /
(1 + self.tyrrell_beta))
new_d[node] = self.clip_a(g, node, new_a)
# TODO Record this in self.flows?
return self.normalize(new_d)
def add_two_gsets(gset1: GSet, gset2: GSet) -> GSet:
'''Combine the gsets, analogous to '+' in the Hopfield equation to
combine two images.'''
result: GSet = defaultdict(set) # type: ignore[arg-type]
edges = union(gset1.keys(), gset2.keys())
for edge in edges:
if edge in gset1:
if edge in gset2:
f = avg_of_funcs(
gset1[edge],
gset2[edge]
)
else:
f = gset1[edge]
else:
f = gset2[edge]
result[edge] = f
return result
def add_two_psets(self, pset1: PSet, pset2: PSet) -> PSet:
'''Combines the psets, analogous to '+' in the Hopfield equation to
combine two images. The default implementation calls .avg_of_funcs()
on the Cartesian product of all the edges (FromTo addresses) in
both psets.'''
result: PSet = defaultdict(set) # type: ignore[arg-type]
edges = union(pset1.keys(), pset2.keys())
for edge in edges:
if edge in pset1:
if edge in pset2:
f = self.avg_of_funcs(pset1[edge], pset2[edge])
else:
f = pset1[edge]
else:
f = pset2[edge]
result[edge] = f
return dict( # remove None funcs
(k, v) for k, v in result.items() if v is not None)
def concentric_walk(self, start_node: Node, dlimit: int) -> Iterable[Node]:
'''Returns a generator of Nodes, starting at start_node, moving out
progressively further hops until reaching dlimit (inclusive).'''
if dlimit >= 0 and self.has_node(start_node):
yield start_node
dist = 1
seen = {start_node}
prev_nodes = {start_node}
while dist <= dlimit:
new_nodes = union(*(self.neighbors(node)
for node in prev_nodes)) - seen
if not new_nodes:
break
yield from new_nodes
# TODO OPTIMIZE Don't make new sets if this is last iteration
seen |= new_nodes
prev_nodes = new_nodes
dist += 1
def on_blocked(self) -> Actions:
'''Actions to perform when this node is Blocked.'''
agents = self.g.neighbors(self, 'agents')
agents_problems = union(*(self.g.neighbors(agent, 'problem')
for agent in agents))
tags_with_agent, tags_without_agent = [], []
for tag in self.g.tags_of(self, 'Blocked'):
if tag in agents_problems:
tags_with_agent.append(tag)
else:
tags_without_agent.append(tag)
if tags_without_agent:
return BuildAgent(self, choice(tags_without_agent))
else:
return BoostFromTo(
intersection(
agents,
self.g.neighbors(tags_with_agent,
neighbor_label='problem')))
def concentric_walk_NodeD(self,
start_node: Node,
dlimit: Optional[int] = None) -> Iterable[NodeD]:
'''Returns a generator of NodeDs, starting at start_node, moving
out progressively further hops until reaching dlimit (inclusive).
If no dlimit is specified, the walk will include all nodes that have
any path from start_node.'''
if self.has_node(start_node) and (dlimit is None or dlimit > 0):
yield NodeD(start_node, 0)
dist = 1
seen = {start_node}
prev_nodes = {start_node}
while True:
new_nodes = union(*(self.neighbors(node)
for node in prev_nodes)) - seen
if not new_nodes:
break
for node in new_nodes:
yield NodeD(node, dist)
seen |= new_nodes
prev_nodes = new_nodes
dist += 1
if dlimit and dist > dlimit:
break
def subprops(self):
if self.kind == PropKind.NAME:
return set()
return set(self.args) | union(arg.subprops for arg in self.args)
nruns=10,
niters=20
):
eqn = eqn * ndups
global gg
gg = make_gset(make_generators(eqn))
for _ in range(nruns):
c = list(startc) * ndups
print(c)
run_gset(c, gg, niters=niters)
print()
if __name__ == '__main__':
eqns = set(make_eqns())
gens = union(*(make_generators(e) for e in eqns))
psa(*gens)
startc = (None, '+', 3, None, 10)
c = list(startc)
print(c)
g = (0, 4, add_n(n=1)) #(4, 2, add_n(n=1))
run_generator(c, g)
print(c)
a, b, f = g
#print(avg_of_funcs('+', '-'))
gset1 = make_gset(make_generators((2, '+', 3, '=', 5)))
psa(*gset1.items())
gset2 = make_gset(make_generators((4, 'x', 6, '=', 24)))
print()
psa(*gset1.items())
gset = add_gsets(gset1, gset2)
def neighbors(self, node: Node) -> Set[Node]:
return union([hop.to_node for hop in self.hops_from_node(node)],
[hop.from_node for hop in self.hops_to_node(node)])
