def example_4():
dt = 0.01
X = np.linspace(0, 30, 600)
sum = np.random.rand(X.shape[0]) / 3
X = 2 * np.sin(2 * np.pi * X * 2.1) + 3 * np.sin(
2 * np.pi * X * 0.3) + 1 * np.sin(2 * np.pi * X * 0.57) * X + sum
#X = X[::2]
p = 50
win = 100
yr = X
X = nn.force_recurrent_gap(X, 10, 3, 0)
print(X.shape)
yn, SSE = NGD_sample_HONU('LNU', X, yr, p, win, 20, 30, 50, 0.05)
divide = 1
subplot(2, 1, 1)
plot(yr * divide, 'k')
plot(yn * divide, 'g')
err = (yr - yn[:yr.shape[0]]) * divide
#err = coarse(err,12,24*4,[0,100],3)
#plot(np.sqrt(err**2),'r')
plot(err, 'r')
subplot(2, 1, 2)
plot(SSE)
#X = np.linspace(0,30,8177*2)
#sum = np.random.rand(X.shape[0])/5
#X = 2*np.sin(2*np.pi*X*2.1)+3*np.sin(2*np.pi*X*0.3)+ sum + 1*np.sin(2*np.pi*X*0.77)+ sum
#plot(X,'y')
show()
def example_2():
X = np.linspace(0, 70, 2400)
sum = np.random.rand(X.shape[0]) / 5
X = 0.5 * np.sin(2 * np.pi * X * 0.4) + 0.3 * np.sin(
2 * np.pi * X * 0.5) # + sum
p = 50
win = 300
yr = X
X = nn.force_recurrent_gap(X, 15, 5, 0)
yn, SSE = batch_MLP('LM', 5, X, yr, p, win, 10, 50, 3, 0.005)
#plot(J.T)
show()
divide = 10
subplot(2, 1, 1)
title('Predikce')
plot(yr * divide, 'k')
plot(yn * divide, 'g')
err = (yr - yn[:yr.shape[0]]) * divide
plot(err, 'r')
subplot(2, 1, 2)
plot(SSE)
title('SSE')
#err = coarse(err,12,24*4,[0,100],3)
#plot(np.sqrt(err**2),'r')
show()
def example_3():
dt = 0.01
X = np.linspace(0, 30, 1000)
sum = np.random.rand(X.shape[0])
X = 2 * np.sin(2 * np.pi * X * 2.1) + 3 * np.sin(
2 * np.pi * X * 0.3) + 1 * np.sin(2 * np.pi * X * 0.77) #+ sum
#X = X[::2]
p = 50
win = 100
yr = X
yr[700:706] = 0
X = nn.force_recurrent_gap(X, 15, 5, 0)
print(X.shape)
#yn = LM_batch('LNU',X,yr,p,win,30,10,10,0.05)
yn, EA, EAP = CGD_batch('LNU', X, yr, p, win, 5, 30, 3)
divide = 1
subplot(3, 1, 1)
plot(yr * divide, 'k')
plot(yn * divide, 'g')
err = (yr - yn[:yr.shape[0]]) * divide
plot(err, 'r')
subplot(3, 1, 2)
plot(EA)
subplot(3, 1, 3)
plot(EAP)
#err = coarse(err,12,24*4,[0,100],3)
#plot(np.sqrt(err**2),'r')
#X = np.linspace(0,30,8177*2)
#sum = np.random.rand(X.shape[0])/5
#X = 2*np.sin(2*np.pi*X*2.1)+3*np.sin(2*np.pi*X*0.3)+ sum + 1*np.sin(2*np.pi*X*0.77)+ sum
#plot(X,'y')
show()
def example_1():
N = 3000
rand = np.random.rand(N)
X = np.linspace(0, 45, N)
sum = np.random.rand(X.shape[0]) / 15
X = 0.5 * np.sin(2 * np.pi * X * 0.4) + 0.3 * np.sin(
2 * np.pi * X * 0.8) + sum
max = np.amax(X)
X = X / max
yr = X
n = 10
X = nn.force_recurrent_gap(X, n, 5, 0)
p = 100
win = 200
yn, SSE = XGD_sample_MLP('NGD', 10, X, yr, p, win, 40, 30, 15, 0.01)
divide = 1
subplot(2, 1, 1)
plot(yr * max, 'k')
plot(yn * max, 'g')
err = (yr - yn[:yr.shape[0]]) * max
plot(err, 'r')
subplot(2, 1, 2)
plot(SSE)
show()
def example_5():
X = np.linspace(0, 30, 1000)
sum = np.random.rand(X.shape[0]) / 3
X = 2 * np.sin(2 * np.pi * X * 2.1) + 3 * np.sin(
2 * np.pi * X * 0.3) + 1 * np.sin(2 * np.pi * X * 0.77) + sum
p = 30
win = 200
yr = X
X = nn.force_recurrent_gap(X, 15, 5, 0)
#print(X.shape)
yn = LM_batch('LNU', X, yr, p, win, 30, 10, 10, 0.05)
#yn,EA,EAP = CGD_batch('LNU',X,yr,p,win,5,30,3)
#yn = LM_batch('LNU',X,yr,p,win,30,10,10,0.05)
divide = 1
subplot(1, 1, 1)
plot(yr * divide, 'k')
plot(yn * divide, 'g')
err = (yr - yn[:yr.shape[0]]) * divide
plot(err, 'r')
show()