iburn estimate can fail

[dflemin3]

traceback:

[ 0.04016588 0.04634279 0.0908399 0.12369704]
Calling MIERUN for run: /Users/Jake/Documents/smart_runs/titan/runmie_titan_test3_zq2rn.scr
SUCCESS: MIERUN run complete: /Users/Jake/Documents/smart_runs/titan/runmie_titan_test3_zq2rn.scr
Calling SMART for run: /Users/Jake/Documents/smart_runs/titan/ap_tmp/runsmart_titan_test3_zq2rn_hitran2012_1666_16666cm.scr
SUCCESS: SMART run complete: /Users/Jake/Documents/smart_runs/titan/ap_tmp/runsmart_titan_test3_zq2rn_hitran2012_1666_16666cm.scr
-12597246.1497
/Users/Jake/Projects/Packages/approxposterior/approxposterior/mcmc_utils.py:39: RuntimeWarning: invalid value encountered in true_divide
result = r/(variance*(np.arange(n, 0, -1)))

ValueError Traceback (most recent call last)
in ()
6 bounds=bounds, which_kernel=which_kernel,
7 n_kl_samples=100000, verbose=False, debug=False,
----> 8 timing=False)
/Users/Jake/Projects/Packages/approxposterior/approxposterior/bp.py in run(self, theta, y, m0, m, M, nmax, Dmax, kmax, sampler, cv, seed, timing, which_kernel, bounds, debug, n_kl_samples, verbose, update_prior, **kw)
285
286 # Estimate burn-in, save it
--> 287 iburn = mcmc_utils.estimate_burnin(sampler, nwalk, nsteps, ndim)
288 self.iburns.append(iburn)
289
/Users/Jake/Projects/Packages/approxposterior/approxposterior/mcmc_utils.pyc in estimate_burnin(sampler, nwalk, nsteps, ndim)
111
112 # Save autocorrelation length
--> 113 autolength.append(np.min(roots))
114
115 # List of chains that we are keeping
/Users/Jake/anaconda2/lib/python2.7/site-packages/numpy/core/fromnumeric.pyc in amin(a, axis, out, keepdims)
2370
2371 return _methods._amin(a, axis=axis,
-> 2372 out=out, **kwargs)
2373
2374
/Users/Jake/anaconda2/lib/python2.7/site-packages/numpy/core/_methods.pyc in _amin(a, axis, out, keepdims)
27
28 def _amin(a, axis=None, out=None, keepdims=False):
---> 29 return umr_minimum(a, axis, None, out, keepdims)
30
31 def _sum(a, axis=None, dtype=None, out=None, keepdims=False):
ValueError: zero-size array to reduction operation minimum which has no identity

potential fix: try/except ValueError loop where iburn set to 1. Add assert statement to ensure sampler has non-zero length, i.e. to ensure than that the MCMC actually ran

[dflemin3]

Current fix: if no roots are found, set iburn to 0

try:
    min_root = np.min(roots)
except ValueError:
    min_root = 0
    warn_msg = "WARNING: Burn-in estimation failed.  iburn set to 0."
    warnings.warn(warn_msg)

autolength.append(min_root)

[dflemin3]

This issue is specific to our current burn-in estimation function, so it's fixed. In #16 I demonstrate that we need to reconfigure how burn-in estimation is done, so future burn-in calculation discussion should go there.