One of the key points of the algorithm is that it is easier to learn to order things than it is to learn a proper metric.
THis is related to the same sort of concept in supervised classifier learning as opposed to learning a recommender. Much previous work started with the observation that learning a correct score would natural provide correct ranking and did not examine whether learning to rank directly might provide a more feasible learning algorithm. The "lots of heavy math" problem that you encountered is not going to go away. There is no way to make sense of machine learning with out the math. Intuitionist approaches do not take you very far down the road because without being able to do the math, you can't tell what is going wrong. On Tue, Nov 23, 2010 at 7:50 AM, gagan chhabra <[email protected]>wrote: > A couple of months ago i posted a link for the research paper I am working > on, which is > based on pairwise similarity learning which being an problem in machine > learning. > > I still dont get the exact meaning what this paper has. > > What I have understood so far is that the paper address to design of a > algorithm which helps in > scaling the pairwise similarity learning of images by using Passive > Aggressive algorithms. > > But the main problem Passive Aggressive algorithms. I Googled it but I got > some results which were again > some papers having some heavy mathematics. > > so please can anyone tell me about them both in brief, what actually i'll > have to do to implement the algorithm > in paper below- > > link for the paper: > > http://static.googleusercontent.com/external_content/untrusted_dlcp/research.google.com/en//pubs/archive/35311.pdf > also what are passive aggressive algorithms!! > > > -- > gagan >
