def processInput(data, notify=default_notify):
# take in the data in the given format and extract all of the fields from it
inputParams = [re.sub(r"\s\s+", " ", str(line.strip().split('=')[1])) for line in data.split('\n')]
A = scipy.mat(inputParams[0])
B = scipy.mat(inputParams[1])
C = scipy.mat(inputParams[2])
tempAdj = scipy.mat(inputParams[3])
AdjMat = tempAdj.T
#num compartments
n = int(inputParams[4])
#num inputs
r = int(inputParams[5])
#num outputs
m = int(inputParams[6])
output = StringIO()
#A = scipy.mat('[1 1 0 0; 1 1 0 0; 1 0 1 1; 0 0 1 1]')
#B = scipy.mat('[1; 0; 0; 0]')
#C = scipy.mat('[1 0 0 0; 0 0 1 0]')
#in adjMat we need to replace all Aii's with the leaks
#AdjMat = A.T
#num compartments
#n = 4
#num inputs
#r = 1
#num outputs
#m = 2
notify(" => Inside input processor!")
# ============================================================================
# === STEP 1. generate the original graph model
# ============================================================================
#after this everything we need is computed, the Jacobian, alphas, betas, etc
myGraphModel = graphModel.graphModel(A, B, C, n, r, m, AdjMat, notify=notify)
notify(" => completed graphModel.graphModel")
# ######
# ##MB Additions/Modifications
# # if myGraphModel.QRank != n:
# # output.write('q rank != n nonobservable')
# # notify('q rank != n nonobservable')
# # notify('process ended')
# # return "model nonobservable"
# #
# # if myGraphModel.RRank != n:
# # output.write('r rank != n noncontrollable')
# # notify('r rank != n noncontrollable')
# # notify('process ended')
# # return "model noncontrollable"
#
#
# if laplaceTools.hasComplexEigenvalues(A):
# output.write('Complex eigenvalues--not all models may be discovered! (Michael Bilow)')
# ## <end> MB Additions/Modifications
if myGraphModel.QRank != n:
output.write('q rank != n nonobservable')
notify('q rank != n nonobservable')
notify('process ended')
return "model nonobservable"
if myGraphModel.RRank != n:
output.write('r rank != n noncontrollable')
notify('r rank != n noncontrollable')
notify('process ended')
return "model noncontrollable"
notify('DONE MAKING ORIG MODEL')
notify(n)
notify(myGraphModel.Rank)
# ============================================================================
# === STEP 2. generate all other graphs, check properties
# ============================================================================
allCandidates = graphTools.generateAllGraphs(myGraphModel.myGraph, n, myGraphModel.Rank, myGraphModel.Rank)
notify('my original model graph (transposed)')
notify( networkx.to_numpy_matrix(myGraphModel.myGraph))
notify('my original model edges')
notify(list(myGraphModel.myGraph.edges()))
#find number of paths from node to observation
numOrigPathsToObs = graphTools.findNumPathsToObs(myGraphModel, C)
#make sure every node has a path from an input, to itself
hasInputConnOrig = graphTools.ensureInputConn(myGraphModel, B)
#find number of paths from the input node to the other nodes
numOrigPathsFromInput = graphTools.findNumPathsFromIn(myGraphModel, B)
#make sure every node has a path from an output, to itself
hasOuputConnOrig = graphTools.ensureOutputConn(myGraphModel, C)
#compare the shortest paths
shortestPathsOrig = graphTools.shortestInOutPaths(myGraphModel, B, C)
origPathsList = []
if len(shortestPathsOrig) > 0:
try:
origPathsList.append([p for p in shortestPathsOrig[0]])
except networkx.exception.NetworkXNoPath:
notify("No Orig Path")
else:
output.write('No path from input to output in the original graph')
output.close()
exit
#find number of traps
numTrapsOrig = graphTools.findNumTraps(myGraphModel)
passGraphCand = []
numPathToObsWrong = 0
numhasInputConnWrong = 0
numCandPathsFromInputWrong = 0
hasOuputConnWrong = 0
numTrapsWrong = 0
numPathsListWrong = 0
notify('number of candidates: %d' % len(allCandidates))
output.write('num of candidates %d' % len(allCandidates))
output.write('\n')
# if we never had any candidates, terminate here
if len(allCandidates) <= 0:
result = output.getvalue()
output.close()
return result
for candidate in allCandidates:
#find number of paths from node to observation
numCandPathsToObs = graphTools.findNumPathsToObs(candidate, C)
#make sure every node has a path from an input, to itself
hasInputConn = graphTools.ensureInputConn(candidate, B)
#find number of paths from the input node to the other nodes
numCandPathsFromInput = graphTools.findNumPathsFromIn(candidate, B)
#make sure every node has a path from an output, to itself
hasOuputConn = graphTools.ensureOutputConn(candidate, C)
#compare the shortest paths
shortestPaths = graphTools.shortestInOutPaths(candidate, B, C)
#find number of traps took 3620 out
numTraps = graphTools.findNumTraps(candidate)
origListResult = (numOrigPathsToObs, hasInputConnOrig, numOrigPathsFromInput, hasOuputConnOrig, numTrapsOrig)
candListResult = (numCandPathsToObs, hasInputConn, numCandPathsFromInput, hasOuputConn, numTraps)
candPathsList = []
try:
candPathsList.append([q for q in shortestPaths[0]])
except networkx.exception.NetworkXNoPath:
a = 1
if numOrigPathsToObs != numCandPathsToObs:
numPathToObsWrong = numPathToObsWrong + 1
if hasInputConnOrig != hasInputConn:
numhasInputConnWrong = numhasInputConnWrong + 1
if numOrigPathsFromInput != numCandPathsFromInput:
numCandPathsFromInputWrong = numCandPathsFromInputWrong + 1
if hasOuputConnOrig != hasOuputConn:
hasOuputConnWrong = hasOuputConnWrong + 1
if origPathsList != candPathsList:
numPathsListWrong = numPathsListWrong + 1
if numTrapsOrig != numTraps:
numTrapsWrong = numTrapsWrong + 1
#paths match
if origPathsList == candPathsList:
#all other results equal
#if numOrigPathsToObs == numCandPathsToObs and hasInputConnOrig == hasInputConn and numOrigPathsFromInput == numCandPathsFromInput:
# if hasOuputConnOrig == hasOuputConnOrig and numTrapsOrig == numTraps:
# passGraphCand.append(candidate)
#if numOrigPathsToObs == numCandPathsToObs:
# passGraphCand.append(candidate)
if origListResult == candListResult:
passGraphCand.append(candidate)
notify('num of candidates left after graph properties checked: %d out of %d' % (len(passGraphCand), len(allCandidates)))
output.write('num of candidates left after graph properties checked: %d out of %d' % (len(passGraphCand), len(allCandidates)))
output.write('\n')
output.write('number of candidates that had incorrect # paths to obs %d' % numPathToObsWrong)
output.write('\n')
output.write('number of candidates that dont have input connectivity %d' % numhasInputConnWrong)
output.write('\n')
output.write('number of candidates that had incorrect # paths from input %d' % numCandPathsFromInputWrong)
output.write('\n')
output.write('number of candidates that dont have output connectivity %d' % hasOuputConnWrong)
output.write('\n')
output.write('number of candidates that have incorrect # traps %d' % numTrapsWrong)
output.write('\n')
output.write('number of candidates that have incorrect shortestpaths %d' % numPathsListWrong)
output.write('\n')
# if we don't have any candidates left, terminate here
if len(passGraphCand) <= 0:
result = output.getvalue()
output.close()
return result
# ============================================================================
# === STEP 3. check alphas, betas
# ============================================================================
numRank = 0
rankFailedCand = None
passLaplaceCand = []
failAlphaLaplaceCand = []
failBetaLaplaceCand = []
for cand in passGraphCand:
#step one rank(A|B) = n
candAB = laplaceTools.makeSymMat(numpy.append(cand.A, B, axis=1))
#if (cand.A == myMatch).all() or (cand.A == myMatch2).all() or (cand.A == myMatch4).all():
# print "--------------FOUND MATCH!!!!!--------------"
# candTF, candCharEqn, candAlphas, candBetas = laplaceTools.calcTF(cand.A, B, C, n)
# print candAlphas
# print candBetas
# print candTF
# print cand.A
# candABRank = laplaceTools.calcRank(candAB)
candABRank = n
if candABRank == n:
#step 2 compare the moment invariants!
candTF, candCharEqn, candAlphas, candBetas = laplaceTools.calcTF(cand.A, B, C, n)
cand.Alphas = candAlphas
cand.Betas = candBetas
#print "Alpha Keys"
currNum = 0
alphasMatch = True
candAlphaKeysList = []
for currDict in candAlphas:
candAlphaKeys = laplaceTools.getOrderedKeys(currDict)
#print candAlphaKeys
#print myGraphModel.alphaKeys[currNum]
candAlphaKeysList.append(candAlphaKeys)
if candAlphaKeys != myGraphModel.alphaKeys[currNum]:
alphasMatch = False
currNum = currNum + 1
#print "Beta Keys"
currNum = 0
for currDict in candBetas:
#print laplaceTools.getOrderedKeys(currDict)
#print myGraphModel.betaKeys[currNum]
currNum = currNum + 1
if alphasMatch:
currNum = 0
candBetaKeysList = []
betasMatch = True
for currDict in candBetas:
candBetaKeys = laplaceTools.getOrderedKeys(currDict)
candBetaKeysList.append(candBetaKeys)
if candBetaKeys != myGraphModel.betaKeys[currNum]:
betasMatch = False
currNum = currNum + 1
if betasMatch:
passLaplaceCand.append(cand)
#print len(passLaplaceCand)
#print 'FOUND CANDIDATE'
#print candTF
#print cand.A
#print candAlphaKeysList
#print candBetaKeysList
notify('the alphas: %s' % str(myGraphModel.alphaKeys))
notify('the betas: %s' % str(myGraphModel.betaKeys))
output.write('the alphas: %s' % str(myGraphModel.alphaKeys))
output.write('\n')
output.write('the betas: %s' % str(myGraphModel.betaKeys))
output.write('\n')
notify('num of candidates left alpha betas checked: %d out of %d' % (len(passLaplaceCand), len(allCandidates)))
output.write('num of candidates left alpha betas checked %d out of %d' % (len(passLaplaceCand), len(allCandidates)))
output.write('\n')
# if we don't have any candidates left, terminate here
if len(passLaplaceCand) <= 0:
result = output.getvalue()
output.close()
return result
# ============================================================================
# === STEP 3. check submatrix jacobians
# ============================================================================
passJacobianRank = []
for cand in passLaplaceCand:
#now calculate the jacobian
cand.calcJacobianRank()
if myGraphModel.Rank == cand.Rank:
passJacobianRank.append(cand)
notify('num of candidates left after calc rank full jacobian: %d out of %d' % (len(passJacobianRank), len(allCandidates)))
output.write('num of candidates left after calc rank full jacobian %d out of %d' % (len(passJacobianRank), len(allCandidates)))
output.write('\n')
# if we don't have any candidates left, terminate here
if len(passJacobianRank) <= 0:
result = output.getvalue()
output.close()
return result
notify("About to compute laplaceTools.reducedJacMat()...")
#first get the simplified jacobian
myGraphModel.Jac = laplaceTools.reducedJacMat(myGraphModel.Jac)
notify("About to compute laplaceTools.calcAllSubranks()...")
#get the ranks
myGraphModel.JacComboRanks = laplaceTools.calcAllSubranks(myGraphModel.Jac, myGraphModel.Rank)
notify("About to process candidates (distributed: %s)..." % DISTRIBUTED)
if DISTRIBUTED:
# compute ranks of all submatrices of the jacobian of the original model
passSubJacobianRankTask = processSingleTotalJacobian.chunks([(myGraphModel, i) for i in passJacobianRank], 10)()
result = passSubJacobianRankTask.get()
passSubJacobianRank = [item for sublist in result for item in sublist if item]
else:
# for now, we'll do it in the non-iterative way
passSubJacobianRank = [x for x in [processSingleTotalJacobian(myGraphModel, i) for i in passJacobianRank] if x is not None]
notify('num of candidates left at the end: %d out of %d' % (len(passSubJacobianRank), len(allCandidates)))
output.write('num of candidates left at the end: %d out of %d' % (len(passSubJacobianRank), len(allCandidates)))
output.write('\n')
# if we don't have any candidates left, terminate here
if len(passSubJacobianRank) <= 0:
result = output.getvalue()
output.close()
return result
output.write('Adjacency graphs\n')
output.write('Original Model\n')
output.write("%s\n" % str(networkx.to_numpy_matrix(myGraphModel.myGraph).T))
output.write('A matrix %s \n' % str(myGraphModel.A))
output.write('\n')
for i, cand in enumerate(passSubJacobianRank):
output.write('Model %d \n' % (i+1))
output.write(str(networkx.to_numpy_matrix(cand.myGraph).T))
output.write('\n')
output.write('A matrix \n%s \n' % str(cand.A))
output.write('\n')
output.write('\n')
result = output.getvalue()
output.close()
return result
def __init__(self, A, B, C, n, r, m, AdjMat, notify=default_notify):
self.A = A
self.B = B
self.C = C
self.n = n
self.r = r
#max number of parameters
self.m = m
self.JacComboCode = None
self.JacComboRanks = None
self.AdjMat = AdjMat
(self.Q, self.QRank) = laplaceTools.calcQ(self.B, self.n, self.A)
notify(" => completed laplaceTools.calcQ")
(self.R, self.RRank) = laplaceTools.calcR(self.A, self.C, self.n)
notify(" => completed laplaceTools.calcR")
#self.betas is a list of dictionaries
#self.alphas is a dictionary of the alphas from the char eqn
self.TF, self.CharEqn, self.Alphas, self.Betas = laplaceTools.calcTF(self.A, self.B, self.C, self.n)
notify(" => completed laplaceTools.calcTF")
#print "my TF: %s" % self.TF
#print "my CharEqn: %s" % self.CharEqn
#print "my Alphas: %s" % self.Alphas
#print "my Betas: %s" % self.Betas
currNum = 0
self.alphaKeys = []
for currDict in self.Alphas:
self.alphaKeys.append(laplaceTools.getOrderedKeys(currDict))
#print 'alphaKeys%d' % currNum
#print laplaceTools.getOrderedKeys(currDict)
currNum = currNum + 1
currNum = 0
self.betaKeys = []
for currDict in self.Betas:
self.betaKeys.append(laplaceTools.getOrderedKeys(currDict))
#print 'betaKeys%d' % currNum
#print laplaceTools.getOrderedKeys(currDict)
currNum = currNum + 1
notify(" => completed laplaceTools.getOrderedKeys")
#need to fix this self.TF = only the numerator now, need to divide each elem by char eqn
self.Jac = laplaceTools.calcJacobian(self.AdjMat, self.Alphas, self.Betas)
notify(" => completed laplaceTools.calcJacobian")
#print self.Jac
self.Rank = laplaceTools.calcRank(self.Jac)
notify(" => completed laplaceTools.calcRank")
#print 'row of orig Jac: %d, col: %d' %(self.Jac.rows, self.Jac.cols)
# self.Rank = 7
#print "done calculating rank stuff"
#print 'AdjMat'
#print self.AdjMat
self.myGraph = self.makeGraph(self.AdjMat)