My Dating Algorithm
November 2nd, 2010One fine day during Summer, I was waiting for caltrain at South Sanfrancisco station. There was a girl near to the ticket machine who was waiting too. She was so cute that I couldn’t wait to tell this to my roommate who was standing beside me. He looked and found her as average looking girl. I became bit upset with his not so affirmative reply
I told, my computer, look how cute and pretty she is. My computer, when fetched her details from its master database, got confused about my reaction as most people in its database had rated her as average looking. It even tried to ask me reevaluate my rating stating the girl’s average ratings. Furious me, announced a penalty to my computer for doubting my judging abilities.
Later I realized that the reason for finding her cute was due to the resemblance of her face to that of my first year college crush. This was one of few secrets which I didn’t tell to my computer and it couldn’t find anywhere else.
The other day I found another girl who looked cute to me and who was looking similar to my one more crush in the final year of college. My computer, as usual, found this girl as average looking again. But previously punished and based on its learning’s from its previous mistakes, it didn’t suggest me to reconsider my rating this time.
In the span of next two months, I found few more girls who were looking similar to my two college crushes in one or other features. One had similar eyes while the other had similar hair. I had told everything about these girls to my computer and what I had liked most in them be it a eyes or hair or smile.
Initially confused computer now had a more clear idea on my taste which is drastically different from the rest of world’s taste. Though the computer didn’t have details about my crushes, it did have details of few girls who I found similar to those crushes. Based on those details, computer analyzed its database and recommended few unseen girls whose details were similar to those of girls that I had found cute.
In fact, the computer went a step further and recommended few more who had combination of multiple features that I had liked most and are individual feature in earlier cases.
My computer, though never aware of my crushes, gave recommendations who look similar to my crushes and sometimes even better a hybrid of both.
PS1: This is how the recommendation of Amazon, Netflix, Linkedin and other sites works.
PS2: The process explained in the above post is called as Collaborative filtering and recommender engines used in the sites mentioned in PS1 is one of examples of collaborative filtering.
PS3: The process of learning itself based on its previous mistakes is called machine learning and plays very important role in A.I. and related research areas.