NNTMM

NNTMM is abbreviation of **N**eural **N**etwork **T**ool for **M**athematical **M**odelling.
The code performs iterative analyses and optimisation with Kohonen maps and feedforward networks.
NNTMM is developed under MATLAB. It has a user-friendly graphical interface where the user can:

- manipulate with input data (sort, select, etc.)
- perform dimension analyses
- clusterize and visualize data using Kohonen maps
- train feedforward network
- perform calculations with trained network
- analyse network calculated data with Kohonen maps
- find optimal parameters

Details of the approach are given in "F.S. Zaitsev. Mathematical modelling of toroidal plasma evolution. - Moscow: MAX Press Publishing Co., 2005, 524 p. The Book" or "F.S. Zaitsev, A.A. Lukianitsa, A.G. Shishkin. Application of Neural Networks to Mathematical Modelling of Toroidal Plasma. Doklady Mathematics. 2003. V. 68, N 3. P. 471-473."

Figures below illustrate some NNTMM screens.

Data analyses and optimization in numerical experiments. SCoPE calculated data using 30 runs.

Data analyses and optimization in numerical experiments. Neural network (trained using SCoPE data ) simulating 20000 SCoPE runs. New optimal sets of parameters are found.