Coefficients of PDE

[Enowyz]

Hi, Lululxvi.
I tried to build a half-half type domain by using two different constant coefficients in one pde, Coefficient changed from 0.3 to 5 when it cross the middle line, like the following graph.

Here is my code:

def main1():    
    def sol_C(x):     
        x1 = np.linspace(-1,1,5) 
        f = np.mat('0.3 0.3 0.3 5 5')
        spline = sp.interpolate.Rbf(x1,f,function='thin_plate',smooth=0, episilon=0)
        result=spline(x)  
        return result
     
    def pde(x, y, x_np): 
        dy_x = tf.gradients(y, x)[0] #T
        dy_x1 = dy_x[:, 0:1]    #Tx
        dy_x2 = dy_x[:, 1:2]    #Ty
        dy_t = dy_x[:, 2:]      #Tt
        dy_x1x1 = tf.gradients(dy_x1, x)[0][:, 0:1]  #Txx
        dy_x2x2 = tf.gradients(dy_x2, x)[0][:, 1:2]  #Tyy
              
        x_np = x[:,0:1]
        cc = sol_C(x_np)
        return ( 
            dy_t  * cc
            - dy_x1x1  - dy_x2x2
        )
    

I got an error about numpy when it calls " cc = sol_C(x_np)":
NotImplementedError: Cannot convert a symbolic Tensor (strided_slice_22:0) to a numpy array.

Also, I think something wrong when I generate my function sol_C, it doesn't output the result cc as I want from 0.3 to 5. Could you please give me some hints for this part?

Thank you! @lululxvi

[lululxvi]

You can not mix-use tf and np. see #185

[lshil00]

Hi, have you solved this problem?
I have the same problem, and my parameter is a function of the output. So I also need to use scipy.interpolate in Tensorflow, but it gave me the error above

[mengxia123]

嗨，你解决了这个问题吗？我有同样的问题，我的参数是输出的函数。所以我还需要在 Tensorflow 中使用 scipy.interpolate，但它给了我上面的错误

嗨，你解决了这个问题吗？我有同样的问题，我的参数是输出的函数。所以我还需要在 Tensorflow 中使用 scipy.interpolate，但它给了我上面的错误

您解决这个问题了吗？能否告知我下，万分感谢