def mulitopt(fitinput, min_, max_, steps, other_guesses, y_sigmas):
better_fits_found = -1
best_fit_quality = 1e10
best_fit_params = []
inital_guesses = np.exp(np.linspace(np.log(min_), np.log(max_), steps))
for guess in inital_guesses:
params, quality = simplex.optimize(*fitinput, y_sigmas=y_sigmas, p0=np.append([guess], other_guesses))
if quality < best_fit_quality:
best_fit_quality = quality
best_fit_params = params
better_fits_found += 1
info("Better Fits found: %s", better_fits_found)
return [best_fit_params, best_fit_quality]
def find_classifier(A, b, c):
n = len(A.D[1])
R_sq = IDs.copy()
for i in IDs:
if n <= len(R_sq):
break
R_sq.add(-i)
print("=== LP: Find vertex ===")
if not simplex.find_vertex(A, b, R_sq):
raise Error("Unable to find start vertex")
print("=== LP: Run simplex ===")
soln = simplex.optimize(A, b, c, R_sq)
w = Vec(FEATUREs, {f: soln[f] for f in FEATUREs})
gamma = soln["gamma"]
return (w, gamma)
import data as d
import helpers as hel
import plot_helpers as plot
import table
import simplex
# Amplifier Bandpass Curve
def gaus(x, mu, sigma, scale, noise_level):
return noise_level + scale * 1 / sqrt(2 * pi * sigma) * exp(
-(x - mu)**2 / 2 / sigma**2)
initial_guess = [36e3, 100, 10, 0.1]
params, quality = simplex.optimize(d.amplifier_f,
d.amplifier_U,
gaus,
p0=initial_guess)
fit = lambda x: gaus(x, *params)
U_max = max(d.amplifier_U)
U_half = U_max / sqrt(2)
f_with_max_U = d.amplifier_f[argwhere(d.amplifier_U == U_max)]
half_U_freq_left = optimize.brentq(lambda x: fit(x) - U_half, 34e3,
f_with_max_U)
half_U_freq_right = optimize.brentq(lambda x: fit(x) - U_half, f_with_max_U,
38e3)
Q_exp = f_with_max_U / (half_U_freq_right - half_U_freq_left)
print("Bandpass Quality:", Q_exp)
plot.plot(d.amplifier_f, d.amplifier_U, "Frequenz in kHz",
"Durchgelassene Spannung in V", "amplifier.pdf", fit)
# print(A.D[0])
features = {'texture(worst)', 'area(worst)'}
A = make_matrix(matutil.mat2rowdict(F), d, features)
b = make_b(A.D[0])
c = make_c(features, F.D[0])
R_square = F.D[0].copy()
n = len(A.D[1])
it = iter(F.D[0])
while len(R_square) < n:
R_square.add(-next(it))
print(A.D[1])
find_vertex(A, b, R_square)
print(R_square)
print('vertex found')
x = optimize(A, b, c, R_square)
pred = Vec(F.D[0], {})
for k, r in matutil.mat2rowdict(F).items():
s = 0
for f in features:
s += r[f] * x[f]
if s > x['gamma']:
pred[k] = 1
else:
pred[k] = -1
print(pred)
print(d)
# print(x['texture(worst)'])
# print(x['area(worst)'])
# print(x['gamma'])
# Fitting refractive indices
def polynom(X, a1, a2):
return [np.sqrt(a1 + a2 * x**2) for x in X]
def laurent(lambda_, a0, a2):
return [np.sqrt(a0 + a2 / x ** 2) for x in lambda_]
# polynom_fit = hel.autofit(spectral_lines, n**2, polynom, p0=(3, 1e-10))
# laurent_fit = hel.autofit(spectral_lines, n**2, laurent)
polynom_fit, polynom_quality = simplex.optimize(spectral_lines, n, polynom, p0=[3, 1e10])
laurent_fit, laurent_quality = simplex.optimize(spectral_lines, n, laurent, p0=[3, 1e10])
plt.clf()
x_messung = spectral_lines
y_messung = n
x_limit = plot.autolimits(x_messung)
x_flow = np.linspace(*x_limit, num=1000)
y_messung = y_messung
plt.plot(x_flow, polynom(x_flow, *polynom_fit), 'g-', label="Fit mit $x^2$")
plt.plot(x_flow, laurent(x_flow, *laurent_fit), 'b-', label="Fit mit $x^{-2}$")
plt.plot(x_messung, y_messung, 'r+', label="Messwerte")
import pprint
import data as d
import helpers as hel
import plot_helpers as plot
import table
import simplex
# Amplifier Bandpass Curve
def gaus(x, mu, sigma, scale, noise_level):
return noise_level + scale * 1 / sqrt(2 * pi * sigma) * exp(-(x - mu)**2 / 2 / sigma**2)
initial_guess = [36e3, 100, 10, 0.1]
params, quality = simplex.optimize(d.amplifier_f, d.amplifier_U, gaus, p0=initial_guess)
fit = lambda x: gaus(x, *params)
U_max = max(d.amplifier_U)
U_half = U_max / sqrt(2)
f_with_max_U = d.amplifier_f[argwhere(d.amplifier_U == U_max)]
half_U_freq_left = optimize.brentq(lambda x: fit(x) - U_half, 34e3, f_with_max_U)
half_U_freq_right = optimize.brentq(lambda x: fit(x) - U_half, f_with_max_U, 38e3)
Q_exp = f_with_max_U / (half_U_freq_right - half_U_freq_left)
print("Bandpass Quality:", Q_exp)
plot.plot(d.amplifier_f, d.amplifier_U,
"Frequenz in kHz", "Durchgelassene Spannung in V", "amplifier.pdf",
fit)
