def testHMCPropagatorMM(self):
mm = InvertibleMatrix(np.array([[1., 0.], [0., 2.]]))
init_state = State(np.random.normal(size=2))
gen = HMCPropagator(self.pdf, self.gradient, self.timestep * 1.5, self.nsteps, mass_matrix=mm)
self.checkResult(gen.generate(init_state, self.nits))
def testHMCPropagatorMM(self):
mm = InvertibleMatrix(np.array([[1., 0.], [0., 2.]]))
init_state = State(np.random.normal(size=2))
gen = HMCPropagator(self.pdf, self.gradient, self.timestep * 1.5, self.nsteps, mass_matrix=mm)
self.checkResult(gen.generate(init_state, self.nits))
def _propagator_factory(self):
gen = HMCPropagator(self.sys.hamiltonian,
self.param.gradient,
self.param.timestep,
self.param.traj_length,
temperature=self.sys.hamiltonian.temperature,
integrator=self.param.integrator,
mass_matrix=self.param.mass_matrix)
return gen
def testHMCPropagator(self):
gen = HMCPropagator(self.pdf, self.gradient, self.timestep,
self.nsteps)
self.checkResult(gen.generate(self.state, self.nits))
def testHMCPropagator(self):
gen = HMCPropagator(self.pdf, self.gradient, self.timestep, self.nsteps)
self.checkResult(gen.generate(self.state, self.nits))
from stuff import numgrad
g = P.gradient(structures=X.ravel())
ng = numgrad(X.ravel(), lambda x: -P.log_prob(structures=X))
print np.max(np.fabs(g - ng))
from csb.statistics.samplers.mc.propagators import HMCPropagator, MDPropagator
from csb.statistics.samplers import State
class MyPDF(object):
def log_prob(self, x):
return P.log_prob(structures=x)
def gradient(self, x, t=0.0):
return P.gradient(structures=x)
pdf = MyPDF()
prop = HMCPropagator(pdf=pdf,
gradient=pdf.gradient,
timestep=0.15,
nsteps=50)
prop = MDPropagator(gradient=pdf.gradient, timestep=0.01)
traj = prop.generate(
State(X.ravel(), np.random.normal(size=X.ravel().shape)), 500, True)
from ensemble_hic.analysis_functions import write_ensemble
X = np.array([
x.position.reshape(n_beads * n_structures,
3).reshape(n_structures, -1, 3) for x in traj
])
write_ensemble(X.reshape(-1, n_beads, 3), '/tmp/out.pdb')
