Hello,
I have a question concerning the number of features that can be included in the training of the support vector machine (svmtrain function).

How many features can be included for one class? 
I have 25 images (16x16=256 features) that I classified (binary). The training procedure gave an accuracy of 71% correctly classified images.
My question is, whether this result is reliable and valid enough. 
In other words, from the mathematical point of view, is there a restriction or trait-off between feature and sample size?
Does it make sense to have more features than samples?  
Thank you very much.
George