Reducing Mean Square Error [closed]

It's difficult to tell what is being asked here. This question is ambiguous, vague, incomplete, overly broad, or rhetorical and cannot be reasonably answered in its current form. For help clarifying this question so that it can be reopened, visit the help center. Closed 10 years ago.

The topic of mine is face recognition using artificial neural networks using MATLAB code.

I tried my work by taking the images开发者_Go百科 from web and started implementing the program by using neural networks.

By typing nprtool command in MATLAB ,i started training, testing and validating the images from database.

When i started to calculate the mean square error for each image ,the values of mse are being increased for each and every training.

I need to decrease the mse.

So is there any body to help me out in telling the command in MATLAB code.

Its also helpful for me if any body provides me the exact MATLAB code for face recognition using artificial neural networks.

---

I don't understand exactly what you want, but I will try my best.

First, as far as I know, there is no command to reduce the mse. I hope the following explanation helps you understand it better.

The goal of the algorithm is to reduce the mse as much as possible (reach a good local minimum), while trying to not overfit the data. This is usually done by using gradient descent. Of course it will not happen that you get a mse of 0, because that would mean your model overfits. Also, the mse values depend on many things, like the data you use (if you use a lot of data it will be harder to create a model that fits the data well, so the mse may be higher). So, I guess, you don't really have a problem with those mse values.

The only command I know which is related with NN and mse is the parameter option, net.performFnc = 'mse', which just tells the algorithm which of all performance functions to use.