Recently the Hawk-Eye technology has been used in tennis to judge close line calls. I had previously posted about Hawk-Eye, thinking at the time that it was the right way to go for tennis. In this post I will try and make a case otherwise :).
The problem with using Hawk-Eye in tennis lies in the fact that what it shows us is the 2d projection of the tennis ball. This is fine when the ball is traveling through the air. But when the ball hits the ground, it is not the 2d projection of the 3d ball which is the point of contact of the ball with the ground. The soft tennis ball gets partially squashed when it hits the ground, & it is the squashed portion of the ball which touches the ground. Thus the point (or region) of contact is actually a fraction of the 2d projection shown to us by Hawk-Eye. Moreover such a fraction would be hard to figure out using simple physics, since it's a chaotic system depending on a lot of factors like speed, spin, angle, surface, temperature etc.
Of course, one can argue that Hawk-Eye need not be perfect, but merely better than the human judges for it to be used. But testing such an assertion would be very tough, simply because there is no ground truth in this problem. Even if such a test comparing Hawk-Eye and human judges were to be performed, who decides whether Hawk-Eye or the human judge is "more right"? That would have to be a system better than both Hawk-Eye or humans. The existence of such a system would imply that we wouldn't have this argument in the first place. Ah paradoxes :).
Showing posts with label technology. Show all posts
Showing posts with label technology. Show all posts
Sunday, July 15, 2007
Monday, March 19, 2007
My blog is a spam
Over the weekend, I found out that Blogger has blocked by blog as spam. I could not add posts anymore.

I had to go through

Good...
Step 1 eliminates all automatically created blogs. Which also reduces the number of supervisors at Google who need to do step 2.
Step 2 eliminates all human generated blogs which are spam
But then the question bugging me is: Why the hell did their algorithm think this blog is a spam? I'm sure they have some probabilistic algorithm to detect splogs based on simple features. But what features do you think this blog has which could characterize it as a spam?
This is a case of a spam detection algorithm throwing up a False Postive. False positives should be a big no-no for any spam/splog detection system. The algorithm error should incline on the side of having some false negatives rather than false positives, for e.g. have some spam show up in your inbox rather than good mail going directly to your junk folder.
All this talk of false positives remind me of the statement that is the basis of judicial systems around the world: "It's better for ten guilty men to go free, than for one innocent man to be convicted.". Same concept, different context.
I had to go through
- Step 1 - a CAPTCHA to prove that I am indeed a human, followed by
- Step 2 - a wait period while some human actually verified that my blog is not spam.
Good...
Step 1 eliminates all automatically created blogs. Which also reduces the number of supervisors at Google who need to do step 2.
Step 2 eliminates all human generated blogs which are spam
But then the question bugging me is: Why the hell did their algorithm think this blog is a spam? I'm sure they have some probabilistic algorithm to detect splogs based on simple features. But what features do you think this blog has which could characterize it as a spam?
This is a case of a spam detection algorithm throwing up a False Postive. False positives should be a big no-no for any spam/splog detection system. The algorithm error should incline on the side of having some false negatives rather than false positives, for e.g. have some spam show up in your inbox rather than good mail going directly to your junk folder.
All this talk of false positives remind me of the statement that is the basis of judicial systems around the world: "It's better for ten guilty men to go free, than for one innocent man to be convicted.". Same concept, different context.
Labels:
AI,
blogs,
spam,
technology
Tuesday, February 06, 2007
Why can't I tag music?
I wonder why songs are not yet 'taggable'. For the uninitiated, a 'tag' is a word or a short phrase describing an entity. For e.g. this blog post is tagged (labelled) with 3 tags: 'Tags', 'Music' and 'Technology' which describe what it talks about. Tags are now ubiquitous in the Web2.0 world. Photo applications like Flickr, Bookmark managers like Del.icio.us, Blog platforms like Blogger and Wordpress etc. use tags. Google is so sold on tags that it did away with folders altogether in Gmail and replaced it with tags (or 'label's as they call it).
Individual songs should be taggable. It would help a lot while playing music which transcends genres. For e.g. many of Eric Clapton songs can be tagged as Classic Rock or Soft Rock or Blues and should be played even if we select either of these 3 genres. Neither iTunes nor MediaMonkey (the app I use to play songs on my laptop) let me do this. Sucks big time.
Individual songs should be taggable. It would help a lot while playing music which transcends genres. For e.g. many of Eric Clapton songs can be tagged as Classic Rock or Soft Rock or Blues and should be played even if we select either of these 3 genres. Neither iTunes nor MediaMonkey (the app I use to play songs on my laptop) let me do this. Sucks big time.
Labels:
music,
tags,
technology
Saturday, February 03, 2007
Cool new way to search
If you are like me, you use the search bar in Firefox at the top right to do your search. If you are like me, you also have a number of search engines which you use pretty frequently e.g. Google, Wikipedia, Weather.com, Amazon etc. etc.
Enter Yubnub. YubNub is a cool engine that allows you to search multiple websites for anything from one place. If defines shortcuts for each search engine e.g. If you want to search "Seattle" in Google all you type is "g Seattle" in YubNub. "wp Seattle" does a Wikipedia search. And it's pretty comprehensive. All the search engines I use, I could find shortcuts for. Just go "ls <search engine name>" to search among existing shortcuts. In the rare case that you can't find a shortcut for your favorite search engine, you can create a new shortcut in minute.
Now all you have to do is add Yubnub as your only Firefox search engine. Then go Ctrl+K and type on. It's gonna save you hundreds of clicks a day.
Enter Yubnub. YubNub is a cool engine that allows you to search multiple websites for anything from one place. If defines shortcuts for each search engine e.g. If you want to search "Seattle" in Google all you type is "g Seattle" in YubNub. "wp Seattle" does a Wikipedia search. And it's pretty comprehensive. All the search engines I use, I could find shortcuts for. Just go "ls <search engine name>" to search among existing shortcuts. In the rare case that you can't find a shortcut for your favorite search engine, you can create a new shortcut in minute.
Now all you have to do is add Yubnub as your only Firefox search engine. Then go Ctrl+K and type on. It's gonna save you hundreds of clicks a day.
Labels:
cool,
technology,
utility,
web,
yubnub
Subscribe to:
Posts (Atom)