Why am I getting Anano TypeError when I try to update a shared variable?

Question

I am trying to run a very simple gradient descent on the function y = x ^ 2. I tried to implement it using the following code:

import theano
from theano import tensor as T
x = theano.shared(2)
y = x ** 2

dy_dx = T.grad(y, x)
learning_rate = 1
updates = [(x, x - learning_rate * dy_dx)]
fn = theano.function([], [y], updates = updates)

But when I try to compile the "fn" function, I get the following error:

TypeError: ('An update must have the same type as the original shared 
variable (shared_var=<TensorType(int64, scalar)>, 
shared_var.type=TensorType(int64, scalar), 
update_val=Elemwise{sub,no_inplace}.0, 
update_val.type=TensorType(float64, scalar)).', 'If the difference is 
related to the broadcast pattern, you can call the 
tensor.unbroadcast(var, axis_to_unbroadcast[, ...]) function to remove 
broadcastable dimensions.')

I thought this could be a problem with the learning_rate variable, as it may not be the same type as the general variable x, but if I change the code as follows:

updates = [(x, x - dy_dx)]

I still get the same error.

I'm stuck :( Any ideas?

Answer 1

, x , . , , Python, , int32. , , dy_dx float64. , , a float64. ​​ ( ), : int32, - float64.

- float. x.

x = theano.shared(2.)