neural network for quadratic programming

Hi,

I'am trying to create a neural network from scratch to solve a quadratic programming

optimization problem with equality constraint.

The neural network is decribed by the equations:

1. 
2. 

where is a variable vector, is a symmetric, positive definite matrix,

is of full row rank, and are constant vectors.

The idea is presented in this paper (Link to IEEE abstract)

Could anyone help me create this program with matlab?

Answer 1

[Greg Heath]

NNs are designed given target/input pairs.

You have no targets.

Or am I missing something???

Hope this helps.

Greg

[Peter Clifford]

Exactly, I asked myself all these questions... The original problem was:

I have considered that the inputs are "chosen" to verify the constraints and that the outputs are determined by any other method available to solve this kind of problem, but in this case what will be the interest of neural networks?

I would like to know if somebody has used neural networks to solve constrained optimization: what should we take as inputs, how to find target(s) (optimal solution), how to program the forward propagation...

etc