def NeuralNet(_data):
_X = XArr.copy()
_y = YArr.copy()
_y = np.reshape(_y, (np.shape(_y)[0], 1))
num_entradas = np.shape(_X)[1]
num_ocultas = 25
numTags = 1
the1 = random_weights(num_entradas, num_ocultas)
the2 = random_weights(num_ocultas, numTags)
theta_vector = np.concatenate((np.ravel(the1), np.ravel(the2)))
start = time.time()
thetas = sciMin(fun = backdrop, x0 =theta_vector, args = (num_entradas, num_ocultas, numTags, _X, _y, lambda_),
method = 'TNC', jac = True, options = {'maxiter': 70}).x
end = time.time()
the1 = np.reshape(thetas[:num_ocultas*(num_entradas+1)], (num_ocultas, (num_entradas+1)))
the2 = np.reshape(thetas[num_ocultas* (num_entradas+1):], (numTags, (num_ocultas+1)))
a,c = evaluateLearning(_y, forward_prop(_X, the1, the2)[0])
print("----------------------------------------------\n")
print("\n TRAINING EXECUTION TIME:", end - start, "seconds")
print("PREDICTION NEURAL_NETWORK: " + str(a) + " %")
models_times[1] = a
print("----------------------------------------------\n")
def doIt(data):
X, y = load_csv(data)
X = np.array(X)
y = np.array(y)
y = np.reshape(y, (np.shape(y)[0], 1))
# ------------------------------
legendPos = np.where(y == 1)
legendX = X[legendPos[0]]
normiePos = np.where(y == 0)
normieX = X[normiePos[0]]
# ------------------------------
num_entradas = np.shape(X)[1]
num_ocultas = 25
num_etiquetas = 1
theta1 = pesos_aleat(num_entradas, num_ocultas)
theta2 = pesos_aleat(num_ocultas, num_etiquetas)
theta_vector = np.concatenate((np.ravel(theta1), np.ravel(theta2)))
thetas = sciMin(fun=backdrop,
x0=theta_vector,
args=(num_entradas, num_ocultas, num_etiquetas, X, y,
lambda_),
method='TNC',
jac=True,
options={
'maxiter': 70
}).x
theta1 = np.reshape(thetas[:num_ocultas * (num_entradas + 1)],
(num_ocultas, (num_entradas + 1)))
theta2 = np.reshape(thetas[num_ocultas * (num_entradas + 1):],
(num_etiquetas, (num_ocultas + 1)))
a, c = checkLearned(y, forward_propagate(X, theta1, theta2)[4])
#b = c
print("Precision de la red neuronal: " + str(a) + " %")
return cost, gradient
X, y = Data_Management.load_mat("ex4data1.mat")
#indexRand = np.random.randint(0, 5001, 100)
#displayData.displayData(X[indexRand[:]])
#plt.show()
weights = loadmat('ex4weights.mat')
theta1, theta2 = weights['Theta1'], weights['Theta2']
theta1 = pesos_aleat(np.shape(theta1)[1]-1,np.shape(theta1)[0])
theta2 = pesos_aleat(np.shape(theta2)[1]-1,np.shape(theta2)[0])
#a1, a2, a3 = propagation(X, theta1, theta2)
theta_vector = np.concatenate((np.ravel(theta1), np.ravel(theta2)))
fmin = sciMin(fun=backdrop, x0=theta_vector,
args=(np.shape(X)[1], 25, 10, X, y, learning_rate),
method='TNC', jac=True,
options={'maxiter': 70})
#backdrop(theta_vector, np.shape(X)[1], 25, 10, X, y, learning_rate)
print(check.checkNNGradients(backdrop, 1e-4))
#print(J(X, transform_y(y, 10), a3, 10, theta1, theta2))
#print(pesos_aleat(5, 6))
weights = loadmat('ex4weights.mat')
theta1, theta2 = weights['Theta1'], weights['Theta2']
theta1 = pesos_aleat(np.shape(theta1)[1] - 1, np.shape(theta1)[0])
theta2 = pesos_aleat(np.shape(theta2)[1] - 1, np.shape(theta2)[0])
num_entradas = np.shape(X)[1]
num_ocultas = 25
num_etiquetas = 10
theta_vector = np.concatenate((np.ravel(theta1), np.ravel(theta2)))
thetas = sciMin(fun=backdrop,
x0=theta_vector,
args=(num_entradas, num_ocultas, num_etiquetas, X,
transform_y(y, num_etiquetas), lambda_),
method='TNC',
jac=True,
options={
'maxiter': 70
}).x
theta1 = np.reshape(thetas[:num_ocultas * (num_entradas + 1)],
(num_ocultas, (num_entradas + 1)))
theta2 = np.reshape(thetas[num_ocultas * (num_entradas + 1):],
(num_etiquetas, (num_ocultas + 1)))
print("Precision de la red neuronal: " +
str(checkLearned(y,
forward_propagate(X, theta1, theta2)[4])) + " %")
#print(check.checkNNGradients(backdrop, 0))
theta2 = pesos_aleat(num_ocultas, num_etiquetas)
theta_vector = np.concatenate((np.ravel(theta1), np.ravel(theta2)))
auxErr = []
auxErrTr = []
thetaMin1 = None
thetaMin2 = None
errorMin = float("inf")
for i in range(1, np.shape(trainX)[0]):
thetas = sciMin(fun=backdrop,
x0=theta_vector,
args=(num_entradas, num_ocultas, num_etiquetas,
trainX[:i], trainY[:i], lambda_),
method='TNC',
jac=True,
options={
'maxiter': 70
}).x
theta1 = np.reshape(thetas[:num_ocultas * (num_entradas + 1)],
(num_ocultas, (num_entradas + 1)))
theta2 = np.reshape(thetas[num_ocultas * (num_entradas + 1):],
(num_etiquetas, (num_ocultas + 1)))
auxErr.append(
J(validationX, validationY,
forward_propagate(validationX, theta1, theta2)[4], num_etiquetas,
theta1, theta2))
auxErrTr.append(
XPoly = Data_Management.add_column_left_of_matrix(XPoly)
XPolyVal = Normalization.normalize2(generate_polynom_data(Xval, 8), mu, sigma)
XPolyVal = Data_Management.add_column_left_of_matrix(XPolyVal)
XPolyTest = Normalization.normalize2(generate_polynom_data(Xtest, 8), mu, sigma)
XPolyTest = Data_Management.add_column_left_of_matrix(XPolyTest)
lambdaAux = [ 0, 0.001, 0.003, 0.01, 0.03, 0.1, 0.3, 1, 3, 10 ]
error_array = np.array([], dtype=float)
error_array_val = np.array([], dtype=float)
thetas = np.array([], dtype=float)
for l in range(len(lambdaAux)):
theta = np.ones(XPoly.shape[1], dtype=float)
theta_min = sciMin(fun=minimizar, x0=theta,
args=(XPoly, y, lambdaAux[l]),
method='TNC', jac=True,
options={'maxiter': 70}).x
error_array = np.append(error_array, J(theta_min, XPoly, y, lambdaAux[l]))
error_array_val = np.append(error_array_val, J(theta_min, XPolyVal, yval, lambdaAux[l]))
thetas = np.append(thetas, theta_min)
lambdaIndex = np.argmin(error_array_val)
plt.figure()
draw_plot(lambdaAux, error_array)
draw_plot(lambdaAux, error_array_val)
plt.show()
theta = np.ones(XPoly.shape[1], dtype=float)
theta_min = sciMin(fun=minimizar, x0=theta,
args=(XPoly, y, lambdaAux[lambdaIndex]),