def spr_test(tree, traits):
#node_opt.bm_brlen_optim(tree,traits,"bfgs")
node_opt.update_all_brlens(tree, traits)
for i in tree.iternodes():
if i.length < 0.001:
i.length = 0.1
print tree.get_newick_repr(True)
cur_best = calc_bm_likelihood.bm_prune(tree, traits)
#rearrange.random_spr(tree)
#cur_best = calc_bm_likelihood.bm_prune(tree,traits)
#rearrange.random_spr(tree)
for i in range(0, 20):
print i
print tree.nnodes("tips")
last_tree = tree.get_newick_repr(True) + ";"
rearrange.random_spr(tree)
#print tree.get_newick_repr(True)
cur_like = calc_bm_likelihood.bm_prune(tree, traits)
print cur_like
if cur_like < cur_best:
print "reverting to previous tree"
tree = tree_reader2.read_tree_string(last_tree)
elif cur_like > cur_best:
cur_best = cur_like
return cur_best
def calc_like_brlens(l, tree, ntraits, sigsq=1.):
bad = tree.update_brlens_all(l)
#bad = tree_utils2.assign_brlens(l,tree)
if bad:
return LARGE
try:
ll = -calc_bm_likelihood.bm_prune(tree, ntraits, sigsq)
except:
return LARGE
return ll
def calc_like_sigsq(sigsq, tree, traits):
if sigsq < 0:
return LARGE
#tree_utils2.assign_sigsq(tree,sigsq)
try:
val = -calc_bm_likelihood.bm_prune(tree, traits, sigsq)
except:
return LARGE
#print ht,val
return val
def calc_like_single_conditional_brlen(bl, node, tree, ntraits, sigsq=1.):
#if bl[0] < 0.00001:
# return LARGE
bad = node.update_child_brlens(bl)
if bad:
return LARGE
try:
ll = -calc_bm_likelihood.bm_prune(tree, ntraits, sigsq)
except:
return LARGE
return ll
def calc_like_brlens(l, tree, traits):
for i in l:
if i < 0:
return LARGE
bad = tree_utils.assign_brlens(l, tree)
if bad:
return LARGE
try:
ll = -calc_bm_likelihood.bm_prune(tree, traits)
except:
return LARGE
return ll
def calc_like_sigsq_brlens(l, tree, ntraits):
bad = tree.update_brlens_all(l[1:])
#bad = tree_utils2.assign_brlens(l,tree)
if bad:
return LARGE
elif l[0] < 0:
return LARGE
try:
ll = -calc_bm_likelihood.bm_prune(tree, ntraits, l[0])
except:
return LARGE
return ll
def calc_like_nodes(ht, tree, traits):
for i in ht:
if i < 0:
return LARGE
bad = tree_utils.assign_node_heights(ht, tree)
#print [i.height for i in tree.iternodes()]
if bad:
return LARGE
try:
ll = -calc_bm_likelihood.bm_prune(tree, traits)
except:
return LARGE
#print ht[0]
#print (val,ht)
return ll
def nni_test(tree, traits):
node_opt.bm_brlen_optim(tree, traits, "bfgs")
#node_opt.update_all_brlens(tree,traits)
#tree.get_newick_repr(True)
for i in tree.iternodes(True):
if i.length < 0:
print i, i.get_newick_repr(True)
cur_best = calc_bm_likelihood.bm_prune(tree, traits)
for i in range(0, 10):
last_tree = tree.get_newick_repr(True) + ";"
rearrange.random_nni(tree)
#rearrange.random_nni_optim(tree,traits)
for node in tree.iternodes():
if node.length < 0:
print node.length, node.get_newick_repr(True)
cur_like = calc_bm_likelihood.bm_prune(tree, traits)
if cur_like < cur_best:
tree = tree_reader2.read_tree_string(last_tree)
elif cur_like > cur_best:
cur_best = cur_like
return cur_best
def calc_like_nodes(ht, tree, traits):
for i in ht:
if i < 0:
return LARGE
z = 0
while z < nrates:
if ht[z] > 500:
return LARGE
z += 1
bad = assign_node_heights(ht, tree)
if bad:
return LARGE
try:
val = -bm_like.bm_prune(tree, traits)
except:
return LARGE
#print ht[0]
#print (val,ht)
return val
def place_fossils(tree, traits, fossil_names, ntraits):
for fossil in fossil_names:
new = node2.Node()
new.label = fossil
new.length = 0.01
new.cont_traits = traits[new.label]
new.istip = True
newpar = node2.Node()
newpar.length = 0.01
newpar.add_child(new)
newpar.cont_traits = np.zeros(ntraits, dtype=np.double)
nodels = [node for node in tree.iternodes() if node != tree]
curbest = -10000000.
for node in nodels:
curlike = 0.
node.parent.add_child(newpar)
node.parent.remove_child(node)
newpar.add_child(node)
newpar.length = node.length / 2
node.length = newpar.length
#estimate brlens for placement and calculate the loglike
calc_bm_likelihood.iterate_lengths(tree, ntraits, 2)
#for n in tree.children:
rr = tree.root_tree()
tree = rr[0]
curlike = calc_bm_likelihood.bm_prune(tree, ntraits)
tree = tree.unroot_tree(rr[1])
print curlike, tree.get_newick_repr(True)
if curlike > curbest:
curbest = curlike
curtree = tree.get_newick_repr(True)
#now restore the original tree and move to the next node
newpar.parent.add_child(node)
newpar.children.remove(node)
newpar.parent.remove_child(newpar)
node.length = node.length * 2
print curtree
def calc_like_sigsq_nodes(ht, tree, traits, nrates):
for i in ht:
if i < 0:
return LARGE
z = 0
while z < nrates:
if ht[z] > 500:
return LARGE
z += 1
if nrates == 1:
tree_utils.assign_sigsq(tree, ht[0])
elif nrates > 1:
tree_utils.assign_sigsq(tree, ht[0:nrates])
bad = assign_node_heights(ht[nrates:], tree)
if bad:
return LARGE
try:
val = -bm_like.bm_prune(tree, traits)
except:
return LARGE
#print ht[0]
#print (val,ht)
return val
def calc_like_strat_bm(p,
tree,
strat,
traits,
nrates=1): #p[0] should be lambda, p[1:nrates] = sig2
for i in p:
if i < 0:
return LARGE
hstart = nrates + 1
if nrates == 1:
tree_utils.assign_sigsq(tree, [p[1]])
elif nrates > 1:
tree_utils.assign_sigsq(tree, p[1:nrates])
bad = tree_utils.assign_node_heights(p[hstart:], tree)
if bad:
return LARGE
try:
sll = -stratoML.hr97_loglike(tree, p[0])
bmll = -calc_bm_likelihood.bm_prune(tree, traits)
ll = sll + bmll
except:
return LARGE
return ll
