############### TRAINING CASES ############### maxIter = 30 xConverge = 0.0005 ## TRAINING CASE 1: nData = (500, 0.2) #data to use, holdback rate nNodes = (100, 30, 30) #hidden nodes, input gran, output gran learning = (0.1, 0.03) #learning rate, momentum val fileName = "FCL_files/DFES_FOMdata_data(%d)_nodes(%d_%d_%d).nwf" % (nData[:1] + nNodes) print "Case MF-2, Triangular: Data Points: %d (%.2f holdback)" % nData, print "Hidden Nodes: %d, Input Granularity: %d, Output Granularity: %d, " % nNodes fuzzData_tri1 = copy.deepcopy(fuzzData_tri[:nData[0]]) sys = DFES(inRanges, outRanges, 'sigmoid', hidNodes=nNodes[0], inGran=nNodes[1], outGran=nNodes[2]) sys.train(fuzzData_tri1, holdback=nData[1], LR=learning[0], M=learning[1], maxIterations=maxIter, xConverge=xConverge) sys.write_weights(fileName) #write network weight foil print "~~~~~~~~~~~~~~~ Optimization Complete ~~~~~~~~~~~~~~~" """ ## TRAINING CASE 2: nData = (500, 0.2) #data to use, holdback rate nNodes = (100, 30, 30) #hidden nodes, input gran, output gran learning = (0.1, 0.03) #learning rate, momentum val fileName = "FCL_files/DFES_FOMdata_data(%d)_nodes(%d_%d_%d).nwf" % (nData[:1] + nNodes) print "Case MF-3, Trapezoidal: Data Points: %d (%.2f holdback)" % nData, print "Hidden Nodes: %d, Input Granularity: %d, Output Granularity: %d, " % nNodes fuzzData_trap1 = copy.deepcopy(fuzzData_trap[:nData[0]]) sys = DFES(inRanges, outRanges, 'sigmoid', hidNodes=nNodes[0], inGran=nNodes[1], outGran=nNodes[2]) sys.train(fuzzData_trap1, holdback=nData[1], LR=learning[0], M=learning[1], maxIterations=maxIter, xConverge=xConverge) sys.write_weights(fileName) #write network weight foil
fileName = "FCL_files/DFES_RFdata_GWT_data(%d)_nodes(%d_%d_%d)_valErr_16Sep15_short.nwf" % (
nData[:1] + nNodes)
print "Case 1-1, Data Points: %d (%.2f holdback)" % nData,
print "Hidden Nodes: %d, Input Granularity: %d, Output Granularity: %d, " % nNodes
print "Learning Rate: %.3f, Momentum: %.3f" % learning
fuzzData = random.sample(fuzzData_GWT, nData[0])
sys = DFES(inRanges,
outRanges_GWT,
'sigmoid',
hidNodes=nNodes[0],
inGran=nNodes[1],
outGran=nNodes[2])
sys.train(fuzzData,
holdback=nData[1],
LR=learning[0],
M=learning[1],
maxIterations=maxIter,
xConverge=xConverge,
combError=False)
sys.write_weights(fileName) #write network weight foil
print "~~~~~~~~~~~~~~~ Optimization Complete ~~~~~~~~~~~~~~~"
## TRAINING Pisnt:
nData = (360, 0.1) #data to use, holdback rate
nNodes = (260, 40, 70) #hidden nodes, input gran, output gran
learning = (0.001, 0.0005) #learning rate, momentum val
fileName = "FCL_files/DFES_RFdata_Pin_data(%d)_nodes(%d_%d_%d)_valErr_16Sep15_short.nwf.nwf" % (
nData[:1] + nNodes)
print "Case 1-1, Data Points: %d (%.2f holdback)" % nData,
# print "Hidden Nodes: %d, Input Granularity: %d, Output Granularity: %d, " % nNodes
print "Learning Rate: %.3f, Momentum: %.3f" % learning
############### TRAINING CASES ############### maxIter = 15 xConverge = 0.005 ## TRAINING GWT: nData = (360, 0.1) #data to use, holdback rate nNodes = (260, 40, 70) #hidden nodes, input gran, output gran learning = (0.001, 0.0005) #learning rate, momentum val fileName = "FCL_files/DFES_RFdata_GWT_data(%d)_nodes(%d_%d_%d)_valErr_16Sep15_short.nwf" % (nData[:1] + nNodes) print "Case 1-1, Data Points: %d (%.2f holdback)" % nData, print "Hidden Nodes: %d, Input Granularity: %d, Output Granularity: %d, " % nNodes print "Learning Rate: %.3f, Momentum: %.3f" % learning fuzzData = random.sample(fuzzData_GWT, nData[0]) sys = DFES(inRanges, outRanges_GWT, 'sigmoid', hidNodes=nNodes[0], inGran=nNodes[1], outGran=nNodes[2]) sys.train(fuzzData, holdback=nData[1], LR=learning[0], M=learning[1], maxIterations=maxIter, xConverge=xConverge, combError=False) sys.write_weights(fileName) #write network weight foil print "~~~~~~~~~~~~~~~ Optimization Complete ~~~~~~~~~~~~~~~" ## TRAINING Pisnt: nData = (360, 0.1) #data to use, holdback rate nNodes = (260, 40, 70) #hidden nodes, input gran, output gran learning = (0.001, 0.0005) #learning rate, momentum val fileName = "FCL_files/DFES_RFdata_Pin_data(%d)_nodes(%d_%d_%d)_valErr_16Sep15_short.nwf.nwf" % (nData[:1] + nNodes) print "Case 1-1, Data Points: %d (%.2f holdback)" % nData, # print "Hidden Nodes: %d, Input Granularity: %d, Output Granularity: %d, " % nNodes print "Learning Rate: %.3f, Momentum: %.3f" % learning fuzzData = random.sample(fuzzData_Pin, nData[0]) sys = DFES(inRanges, outRanges_Pin, 'sigmoid', hidNodes=nNodes[0], inGran=nNodes[1], outGran=nNodes[2])
