Thursday, December 08, 2005,10:01 AM
PR project... pain in the a**
Nice day today. I got up at 9 AM. My first reaction was, what the hell, CV class again gone! Two classes in a row! Man, got to do something about it now. Anyways, Srikanth did give me a proxy. I hope that works. Then, at about 10:30 AM, I came to know that I have to leave for Visu at 12. I had to register my BTP too. Nikhil had a class at 11:30, and Srikanth was down with fever. I hurriedly got ready and rushed to Prof. Jawahar.

We registered for a BTP, as one team, a team of three. But, sir suggested to do it 2+1 or 1+1+1 instead of all three in one team. We waguely agreed. Lets see what happens. Sir also told me that I had been selected for the DS TAship, and I should report for work as soon as possible. Hurray! He asked me why did I miss two classes in a row. Oh my god! Does he know that I didn't come today? Does he know that I was absent today? Ofcourse he does! Did he mark me absent then? Anyways, lets hope that I get the attendance for today.

Oh, and by the way, I chose Prof. Jawahar finally, after one night of dilemma. I guess we'll do quite well under him. Lets hope for the best.

We went to Visu today. Lakshmi and Bhanu couldn't make it due to I don't know what! Anyways, we reached quite very much in time there. We finally enrolled for the classes. Thursdays and Sundays, a total of 8 hours a week and 4 hours of self study per day. It looks to be tough. I hope I can do it.

I called up Riya when I was in Visu. Her DIP paper was nice. Most of it was morphological things. And she was like, bye bye bye, keep down the phone monkey. And I was not doing it. We kept on quarreling on keeping down the phone. Finally, she gave up and kept it down, as in disconnected it, hehe. I win, I win, I win!

Okay fine, this PR project is really pissing me off. We tried to see if that blurred smooth image can give us some output. I thought, maybe the training is less. We tried increasing epochs. Its going around 7-8% error till about one lakh epochs. I tried 8 lakh epochs, and it overfitted to give 20% error. I guess I will first make it display what it means by 7% error. If it has any problems, we might change the hidden nodes. Hope things work.

--- continued after a break ---
At 12 AM, Neeraj called me up. It was a BC meeting as usual. Nice friendly talk, teasing and stuff. I have just returned. Its damn cold right now. I am almost frozen. I will now do the PR project work. Tanu has a paper tomorrow, rather today. Hope she does well.
 
posted by Anurag Singh Rana
Permalink ยค


0 Comments: