def obj_func(sampler, distr, job_id, **kwargs):
num_target_grad_evals = grad_evals[type(distr).__name__]
default_args = {
"num_grad_steps": num_target_grad_evals,
"sample_steps": 1,
"num_steps": None,
"half_window": True
}
kwargs = unpack_params(kwargs)
kwargs.update(default_args)
ac_df = calculate_autocorrelation(sampler, distr, **kwargs)
n = ac_df['num grad'].values.astype(int)
# necessary to keep curve_fit from borking
normed_n = n / (0.5 * num_target_grad_evals)
y = ac_df['autocorrelation'].values
if use_exp:
# wtf fit is mutating input somehow
r1 = fit(normed_n.copy(), y.copy())[0]
if debug:
plot_fit(normed_n, y, r1, job_id, kwargs)
return -r1
else:
if np.isnan(np.sum(ac_df.autocorrelation.values)):
return 11 * num_target_grad_evals
for trial, target in enumerate(np.arange(0, 1, 0.1)):
score = min_idx(ac_df, target)
if score is not None:
score += trial * num_target_grad_evals * 0.5
break
if debug:
plot_search_ac(ac_df['num grad'].values.astype(int),
ac_df.autocorrelation.values,
job_id,
kwargs, score)
return score or 5 * num_target_grad_evals
def obj_func(sampler, distr, job_id, **kwargs):
""" Scores the performance of sampler on distribution given parameters
:param sampler: sampler being tested. instance of mjhmc.samplers.markov_jump_hmc.HMCBase
:param distr: distribution being used. instance of mjhmc.misc.distributions.Distribution
:param job_id: integer label for job being run
:returns: the score
:rtype: float
"""
cos_coef, normed_n_grad_evals, exp_coef, autocor, kwargs = obj_func_helper(sampler, distr, True, kwargs)
if debug:
plot_fit(normed_n_grad_evals, autocor, exp_coef, cos_coef, job_id, kwargs)
return exp_coef
def obj_func(sampler, distr, job_id, **kwargs):
""" Scores the performance of sampler on distribution given parameters
:param sampler: sampler being tested. instance of mjhmc.samplers.markov_jump_hmc.HMCBase
:param distr: distribution being used. instance of mjhmc.misc.distributions.Distribution
:param job_id: integer label for job being run
:returns: the score
:rtype: float
"""
cos_coef, normed_n_grad_evals, exp_coef, autocor, kwargs = obj_func_helper(sampler, distr, True, kwargs)
if debug:
plot_fit(normed_n_grad_evals, autocor, exp_coef, cos_coef, job_id, kwargs)
return exp_coef
def plot_all_best(custom_params=None):
""" Creates a plot with the autocorrelation and fit for each distribution and sampler
:param custom_params: dictionary of custom params will be used on all distributions
and samplers. if None uses the current best params for each
:returns: None
:rtype: None
"""
distributions = [
RoughWell(nbatch=200)
# Gaussian(ndims=10, nbatch=200),
# MultimodalGaussian(ndims=5, separation=1)
]
samplers = [
ControlHMC,
MarkovJumpHMC
]
with PdfPages("validation.pdf") as pdf:
for distribution in distributions:
# [control, mjhmc, lahmc]
if custom_params is None:
params = load_params(distribution)
else:
params = [custom_params] * 3
active_params = params[:-1]
for sampler, hparams in zip(samplers, active_params):
print "Now running for {} on {}".format(sampler, distribution)
cos_coef, n_grad_evals, exp_coef, autocor, _ = obj_func_helper(
sampler, distribution.reset(), False, hparams)
fig = plot_fit(n_grad_evals,
autocor,
exp_coef,
cos_coef,
'validation',
hparams,
save=False
)
pdf.savefig(fig)
