def test_random_model(self):
num_segments = 5
num_clones = 3
num_alleles = 2
model = self.create_random_model(num_segments, num_clones, num_alleles)
x = np.random.randint(100000, size=(num_segments, 3))
for m in xrange(num_clones):
for ell in xrange(num_alleles):
elbo_1 = model.calculate_elbo(x)
model.update(x)
elbo_2 = model.calculate_elbo(x)
print elbo_2 - elbo_1
self.assertTrue(elbo_2 - elbo_1 > -1e-10)
def test_update_swap(self):
num_segments = 4
for i in xrange(num_replicates):
cn = np.array([[[1, 1], [0, 1]], [[1, 1], [2, 1]], [[1, 1], [0,
1]],
[[1, 1], [0, 1]]])
h = np.array([0.2, 0.3])
l = np.array([10000., 30000., 20000., 10000.])
phi = np.array([0.2, 0.2, 0.2, 0.2])
mu = remixt.likelihood.expected_read_count(l, cn, h, phi)
x = np.array([np.random.poisson(a) for a in mu])
x.sort(axis=1)
x = x.reshape(mu.shape)
model = remixt.model3.RemixtModel(
2,
num_segments,
1,
cn_max,
x,
np.array([0, 0, 0, 1]),
np.array([0, 0, -1, -1]),
np.array([-1, +1, 0, 0]),
l,
l,
1.0,
)
print h
model.prior_variance = 1e5
model.h = np.array([0.2, 0.3])
model.init_p_cn()
print model.get_cn()
# model.posterior_marginals[0, :, :, :] = (
# np.array([[[0.01, 0.99],
# [0.01, 0.99],
# [0.01, 0.99],
# [0.01, 0.99]],
# [[0.01, 0.99],
# [0.01, 0.99],
# [0.01, 0.99],
# [0.01, 0.99]]]))
elbo_prev = None
for i in xrange(20):
elbo = model.update()
print 'elbo', elbo
print 'h', model.h
if elbo_prev is not None:
print 'diff:', elbo - elbo_prev
self.assertTrue(elbo - elbo_prev > -1e-5)
elbo_prev = elbo
print model.get_cn()
print np.asarray(model.p_allele)
for n in range(2):
for v in range(2):
for w in range(2):
print n, v, w, model.p_allele[n, v, w]
print np.asarray(model.p_breakpoint)
brk_cn = np.argmax(model.p_breakpoint, axis=-1)
print brk_cn
self.assertTrue(np.all(brk_cn == np.array([0, 1, 0])))
def test_update(self):
num_segments = 4
for i in range(num_replicates):
cn = np.array([[[1, 1], [1, 1], [1, 1]], [[1, 1], [1, 0], [1, 1]],
[[1, 1], [1, 1], [1, 1]], [[1, 1], [0, 0], [0, 0]]])
h = np.array([0.2, 0.3, 0.1])
l = np.array([10000., 30000., 20000., 10000.])
phi = np.array([0.2, 0.2, 0.2, 0.2])
mu = remixt.likelihood.expected_read_count(l, cn, h, phi)
x = np.array([np.random.poisson(a) for a in mu])
x = x.reshape(mu.shape)
model = remixt.model3.RemixtModel(
3,
num_segments,
1,
cn_max,
x,
np.array([0, 0, 0, 1]),
np.array([0, 0, -1, -1]),
np.array([+1, -1, 0, 0]),
l,
l,
1.0,
)
print(h)
model.prior_variance = 1e5
model.h = np.array([0.2, 0.3, 0.1])
print(np.asarray(model.cn_states))
model.init_p_cn()
print(model.posterior_marginals)
print(model.get_cn())
# model.posterior_marginals[0, :, :, :] = (
# np.array([[[0.01, 0.99],
# [0.01, 0.99],
# [0.01, 0.99],
# [0.01, 0.99]],
# [[0.01, 0.99],
# [0.01, 0.99],
# [0.01, 0.99],
# [0.01, 0.99]]]))
elbo_prev = None
for i in range(20):
elbo = model.update()
print('elbo', elbo)
print('h', np.asarray(model.h))
if elbo_prev is not None:
print('diff:', elbo - elbo_prev)
self.assertTrue(elbo - elbo_prev > -1e-5)
elbo_prev = elbo
print(model.get_cn())
print(np.asarray(model.p_garbage))
print(np.argmax(model.p_garbage, axis=-1))
brk_cn = np.argmax(model.p_breakpoint, axis=-1)
self.assertTrue(np.all(brk_cn == np.array([0, 1, 0])))