How to include a pattern for'unknown' for an SVM classifier? I am doing a classification of heart beat with SVM. There are five kinds of beats in my training data. I plan to add a new kind of data named 'unknown' beat. If there is no unknown beat, one to-be-test beat will always be classified into one of the five beats even though it is far from these five beats. That's why I want to have a sixth beat called 'unknown'. Then I found posterior probability seems to help me with it. But When working into it, I didn't find a good way to deal with posterior probability. Hope some one could give me some hints to solve this problem. 
edit:
Actually there are over sixteen kinds of heart beats in all. But there are five main kinds of beats which occupy nearly 90% of all the beats. My work is to classify the heart beats, and if some abnormal beats or some beats related to some disease is found, there beats will be sent to the doctor for further work. The 'unknown' beat is a set of the remaining 11 kinds of beat(except the five main kinds of beats).
 A: Basically you want to use that sixth class as a junk class, right? Everything that does not look like a heartbeat put it there. I would try to adopt a two stage approach:
(1) One-class SVM to separate beats from non-beats
(2) For all beats patterns, do one-vs-all SVM classifier (or any other multiclass classifier).
By (1) you actually model your class of interest (which are all the five beats classes). This classifier should help you to get rid of those bad patterns. The (2) gets only clean patterns that need to be classified in one of five classes.
A: If you don't mind me asking, why use a SVM? why not do something like a multi-class logistic regression (apologies if you have already tried it but thought I should put this in there). 
As a tip, this is a scenario very similar to what you use in Natural Language Processing (NLP) so you might think of borrowing a lot of the results developed there. For e.g., \ to classify words in a sentence what you suggested is essentially softwares you do for Proper Nouns (e.g. James). These words either (a) have not occurred in the sentence or (b) are words your machine did not find in a dictionary (so they are put in an `other' bucket)
