[theano-users] Changing cost function depending on class

[Sid]

I am trying to optimise a cost function where I find the mean Euclidean
distance between each data point and a weight. However, depending on the
class of the data, I use the weight for that particular class. So the cost
function looks like:
sum_i ||x_i - w_{y_i}||^2

How do I implement this cost function in Theano?

Thanks