How to make the predicted value non trivial？

[chuanhon]

The question is -u''=λu, λ is unkown, and 0<x<pi, and u(0)=u(pi)=0.
Code as follow,
import deepxde as dde
import matplotlib.pyplot as plt
import numpy as np

def pde(x, y):
dy_xx = dde.grad.hessian(y, x)
return -dy_xx - y

def boundary(x, _):
return np.isclose(x[0], 0) or np.isclose(x[0], np.pi)

def func(x):
return np.sin(x)

test_train=np.random.uniform(0,np.pi,50)
TestTrain=test_train.reshape(50,1)

geom = dde.geometry.Interval(0, np.pi)
bc = dde.icbc.DirichletBC(geom, func, boundary)
data = dde.data.PDE(geom, pde, bc, 50, 2, anchors=TestTrain, solution=func, num_test=100)

layer_size = [1] + [50] * 3 + [1]
activation = "tanh"
initializer = "Glorot uniform"
net = dde.nn.FNN(layer_size, activation, initializer)

model = dde.Model(data, net)
model.compile("adam", lr=0.001, metrics=["l2 relative error"])
losshistory, train_state = model.train(iterations=10000)

dde.saveplot(losshistory, train_state, issave=True, isplot=True)

see,it predicted 0

[haison19952013]

You can refer to this one: #356 (comment)...

[chuanhon]

thanks.It do help

[pescap]

Hi,
You are solving:

-dy_xx - y = 0 
y(0) = y(\pi) = 0

This problem has no source (or forcing) term and boundary data are 0. The solution to your problem as it is posed is y=0.