Need compute-intensive calculations in your desktop applications. And don't want to write entire DSP or other complex codes in VB/C#/Java? Well there is a solution.

One of my colleagues wanted to write a Neural Network (NN) Application in a project. Because it was meant for distribution, the GUI had to be pretty as well. Obviously, when it comes to visual appearance, nobody likes MATLAB. So he developed the UI in Visual C#, and used advanced controls on the Forms.

Next, using MATLAB Interface to .NET Framework, he implemented the neural network model (which he designed by himself fully optimized for his specific application) and interfaced it with his .Net application.

As an example, the Fast Fourier Transform (FFT) code was one (small!) part of his NN. Imagine the complexity reduced by avoiding custom code/library interfacing. Moreover, MATLAB's ability to manipulate arrays outperforms that of C# or other programming languages.


The downside... Well, now there are two dependencies.
The first is the .Net Runtime itself. The second is the MATLAB Runtime, which also needs to be deployed.