algorithms based on probability theory and statistics

[tubybugydum]

Hello! Sorry for my english.
I have to write graduate work, I want to learn something dealing with neural network or genetic algorithm, or with image processing. The only one limitation is the algorithm must based on probability theory or math statistic and it should be rather new.
Does anybody know some interesting thing?

[MeTh0Dz]

You should not be trying to get input on your graduate thesis from a programming forum.

[Predictor]

Hello! Sorry for my english.
I have to write graduate work, I want to learn something dealing with neural network or genetic algorithm, or with image processing. The only one limitation is the algorithm must based on probability theory or math statistic and it should be rather new.
Does anybody know some interesting thing?

Some neural networks (with softmax outputs, for instance) will estimate probabilities directly. Others estimate likelihoods which approximate probabilities and could be easily curve-fit afterward to produce probability estimates. I wrote an article about pixel classification which you can find here:

Pixel Classification Project

It was in MATLAB, and I used a logistic regression, but that is very similar to a single artificial neuron like a perceptron. Anyway, there is no reason that the same basic process could not be done using neural networks.