Pybrain implement network whose outputs are parameters of a function

Question

I want to build a neural network in pybrain to fit the following function

z(t) = A + B x(t) + C(t) y(t)

Here, (x, y, z)(t) are my experimental data; A and B are constants and C(t) is a function of time. The requirements for my network are:

• Input layer of 4 neurons: t, x, y and z
• output layer of 3 neurons: A, B and C

The two conceptual problems that prevent me from implementing it are

• To train the network, rather than providing the input and output values, I want to minimize the above function
• As I have some prior knowledge (i.e., A and B are constants), I want to explicitly make that they are not time dependent

If the solution is easier using other tools (pylearn2 or scikit-learn) I could also implement it there.

Answer 1

关于第二个问题，pybrain.structure.FullConnection有 kwargs inSliceFrom和inSliceTo设置将要采取的神经元。