def fuzErrorAC(dataF, outF, acuts=11):
#test fuzzy error function based on interval math on alpha cuts
#get fuzzy sets for outF (x,y) and dataF (x,y) and get error using acuts # of cuts
#dataF -
#outF -
#acuts - number of alpha cuts to compare system at (default = 11) -
ac = np.array(range(acuts))/float(acuts-1) #set alpha cuts
#get lambda cuts for data and output
dataLC = [fuzz.lambda_cut(dataF[1], c) for c in ac]
outLC = [fuzz.lambda_cut(outF[1], c) for c in ac]
#get alpha cut intervals for each MF from lambda cuts
dataACi = []
outACi = []
for i in range(len(dataLC)):
dataACi.append( [j for j, x in enumerate(dataLC[i]) if x == 1])
for i in range(len(outLC)):
outACi.append( [j for j, x in enumerate(outLC[i]) if x == 1])
dataAC = []
outAC = []
for i in range(len(dataACi)):
if len(dataACi[i]) > 0:
dataAC.append([ dataF[0][min(dataACi[i])],
dataF[0][max(dataACi[i])] ])
else: dataAC.append([])
for i in range(len(outACi)):
if len(outACi[i]) > 0:
outAC.append([ outF[0][min(outACi[i])],
outF[0][max(outACi[i])] ])
else: outAC.append([])
#get differences in intervals
error_ints = []
for i in range(len(dataAC)):
if len(dataAC[i]) > 0 and len(outAC[i]) > 0:
error_ints.append([ min(dataAC[i]) - min(outAC[i]),
max(dataAC[i]) - max(outAC[i]) ])
#get the position error (do the fuzzy sets line up)
if len([sum(err)/len(err) for err in error_ints]) > 0.0:
pos_error = sum([sum(err)/len(err) for err in error_ints])/ \
len([sum(err)/len(err) for err in error_ints])
else: pos_error = 100.0
#get the spread error (do the sets have the same MF shape)
if len([(err[0])+(err[1]) for err in error_ints]) > 0.0:
spread_error = sum([(err[0])+(err[1]) for err in error_ints])/ \
len([(err[0])+(err[1]) for err in error_ints])
else: spread_error = 100.0
pos_sign = 1
if pos_error < 0: pos_sign = -1 #get sign of error
return pos_error, spread_error
def alpha_at_val(x,y,alpha=None):
"""
Takes in a fuzzy membership function, determines, the maximum membership degree,
and then returns the alpha-cut interval at that maximum
------- INPUTS ------
x : list/array
x values for membership function
y : list/array
y values for membership function
alpha : float
alpha value to find alpha-cut at (If None, uses max of MF)
"""
try:
if alpha == None: alpha = 0.999*max(y) #get maximum membership
lc = fuzz.lambda_cut(y, alpha) #get lamda cut
mm = [x[i] for i in range(len(lc)) if lc[i] == 1] #get x values for alpha cut
if not isinstance(mm, list): mm = [mm]
if len(mm) > 0: #get range of cut
return [ min(mm), max(mm)]
elif alpha == None: #catch for single peak value
y = list(y)
imax = y.index(max(y))
print imax
return [x[imax], x[imax]]
else: #if cut is empty
return [0, 0]
except:
return [None, None]
def test_lambda_cut():
x = np.arange(21) - 10
mfx = fuzz.trimf(x, [-2, 3, 5])
# fuzz.lambda_cut test
expected = np.r_[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0,
0, 0]
result = fuzz.lambda_cut(mfx, 0.33)
assert_allclose(expected, result)
# fuzz.arglcut test
expected = np.r_[10, 11, 12, 13, 14]
result = fuzz.arglcut(mfx, 0.33)
assert len(result) == 1
assert_allclose(expected, result[0])
def test_lambda_cut():
x = np.arange(21) - 10
mfx = fuzz.trimf(x, [-2, 3, 5])
# fuzz.lambda_cut test
expected = np.r_[0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0]
result = fuzz.lambda_cut(mfx, 0.33)
assert_allclose(expected, result)
# fuzz.arglcut test
expected = np.r_[10, 11, 12, 13, 14]
result = fuzz.arglcut(mfx, 0.33)
assert len(result) == 1
assert_allclose(expected, result[0])
