A neural network is a mathematical technique based on the natural neural network in the human brain.
To interpolate a function, a network with three levels (Inp, Hidden, Out) is built and the connections between them are weighed.

Each arrow is associated with a weight () and each ring is called a cell (like a neural).
If the inputs are , the hidden level contains function
and the output solution is also:
where is a predefined function, such as the hyperbolic tangent or an exponential
based function, to obtain something similar to the binary behavior of the electrical brain
signal (like a step function). The function is continuous and differentiable.
The weight functions () are issued from an algorithm that minimizes (as the least squares method) the distance between the interpolation and the known values (design points). This is called the learning. The error is checked at each iteration with the design points that are not used for the learning. Learning design points need to be separated from error-checking design points.
The error decreases and then increases when the interpolation order is too high. The minimization algorithm is stopped when the error is the lowest.
This method uses a limited number of design points to build the approximation. It works better when the number of design points and the number of intermediate cells are high. And it can give interesting results with several parameters.