#INative = np.array(mInfo.ContactStateIndices)==mInfo.NativeContactStateIndex
#stability = np.real( EigAns[1][INative,0]/EigAns[1][:,0].sum() )
stability = 0.0 # for testing
# print stability and timescale info
print '%d\t%8.3f\t%16.8f\t%16.8f'%(TotalCounts, stability, result[0,1], result[1,1]),
else:
print '%d\tnot enough counts yet...'%(TotalCounts),
# revise our estimates of how many samples to take from each state
if Adaptive:
lagTime = 1
TotalSamples = sum(NumSamplesPerState) # this is the number of new samples we want to make each time'
#print 'TotalSamples', TotalSamples
s = ssaCalculator(lagTime, C.todense(), nNewSamples=TotalSamples, evalList=[1,2], recommendationScheme = 'Nina')
#print 's.varianceContributions', s.varianceContributions
for EigInd in range(len(s.evalList)):
#print s.uncertainty_variance(EigInd),
if EigInd == 0:
RecommendedSamplesPerState = list(s.resamplingDistribution(EigInd))
NumSamplesPerState = RecommendedSamplesPerState
if (0):
print
print 's.weights', np.array(s.weights)
print 's.qlists[EigInd=0]', s.qlists[0]
print 's.resamplingDistribution', 'for eigenvalue index', s.evalList[0], repr(RecommendedSamplesPerState).replace(' ','').strip()
print '----------'
sys.exit(1)
# Find all the non-ergodic states and sample only these in the next round
I_nonergodic = (Mapping == -1)
MacroNumSamplesPerState = np.where(I_nonergodic, nsamples_uniform, 0)
NumSamplesPerState = [MacroNumSamplesPerState[mInfo.ContactStateIndices[i]] for i in range(NumStates)]
if C_trimmed.shape == C_macro.shape:
ErgodicMacro = True
print 'C_macro', C_macro
# Next, use Nina's adaptive sampling algorithm to determine which macrostate should be sampled at each round
for trial in range(NumTrials):
lagTime = 1.
s = ssaCalculator(lagTime, C_macro.todense(), evalList=[1,2], recommendationScheme = 'Nina', nNewSamples=nsamples_adaptive)
EigInd = 0
MacroNumSamplesPerState = np.array(s.resamplingDistribution(EigInd)) # Recommended Samples
print 'Adaptive sampling macrostate indices', MacroNumSamplesPerState.nonzero()[0]
NumSamplesPerState = [MacroNumSamplesPerState[mInfo.ContactStateIndices[i]] for i in range(NumStates)]
# sample microstates corresonding to the recommended macrostate
for i in np.array(NumSamplesPerState).nonzero()[0]:
if i%1000 == 0:
print 'Sampling', NumSamplesPerState[i], 'transitions from state', i, 'of', NumStates, '...'
# look for possible transitions
possible_j = jumps[i][0]
jprobs = jumps[i][1]
# sample counts from state i
picks = draw_index(jprobs, n_picks=NumSamplesPerState[i])
# print stability and timescale info
print '%d\t%8.3f\t%16.8f\t%16.8f' % (TotalCounts, stability,
result[0, 1], result[1, 1])
else:
print '%d\tnot enough counts yet...' % (TotalCounts)
# revise our estimates of how many samples to take from each state
if Adaptive:
lagTime = 1
TotalSamples = sum(
NumSamplesPerState
) # this is the number of new samples we want to make each time'
# print 'TotalSamples', TotalSamples
s = ssaCalculator(lagTime,
C,
nNewSamples=TotalSamples,
evalList=[1, 2],
recommendationScheme='Nina')
#print 's.qlist', s.qlist
#print 's.varianceContributions', s.varianceContributions
for EigInd in range(len(s.evalList)):
print s.uncertainty_variance(EigInd),
#print 's.resamplingDistribution', 'for eigenvalue index', s.evalList[EigInd]
RecommendedSamplesPerState = list(s.resamplingDistribution(EigInd))
print np.argmax(RecommendedSamplesPerState)
if EigInd == 0:
NumSamplesPerState = RecommendedSamplesPerState
sys.exit(1)