Today, I read over the multiclassification paper again, as well as Sal's header file. I have some questions about the way in which his code works, but I will have to ask him later this week as I didn't make it over to the HITLab this morning. I don't think that matters too much because he sounded like he was rather busy today.

For lunch, I played bridge again with the CS grad people. What I have learned from that is I should get or borrow a book on bridge.

This afternoon, I met with Raj Rao to discuss grad school options, as well as options for this next year. Our discussion was highly informative--he gave me many names of people whom I should contact. One, in particular, is Oren Etzioni who got some grants to work on Natural Language Processing. I should email him this week.

To do:

1. Look up more information on support vector machines so that Sal's code might make some more sense.
2. Reread first paper on multiclassification.
3. Start coding up length-angle histograms as we described in Friday's meeting (also, figure out how to display the results graphically, sort of as a spin image diagram).
4. Figure out what a bilinear interpolation is.
5. Put together notes I made on ideas for approach and talk with Linda and/or Sal about them.

Arati and I went to Vas's thesis defense today. It was interesting to see how the committee interacted with the presentor. All in all, it seemed to go well and Vas passed. So, now he just needs to finish up the official report and he is then on his way.

In the afternoon, I spoke with Rachel Pottinger about grad school stuff. She directed me to an article she wrote about applying to grad school, which is linked from her homepage (cs./~rap). It was comforting to talk with her. She also said that I should speak with Peter Mork, given my interests, and she sent an email introducing the two of us.

Today I worked on getting gnuplot installed on the HITLab machine, so that I could output a graph for the length-theta signatures. I also looked at how to interface gnuplot with C++. I have yet to figure out exactly what is meant by a bilinear interpolation--I have looked online, but to little avail. I have emailed Sal, asking him if we are to meet as a group or individually tomorrow. I should be able to progress more quickly after touching base with him.

Earlier in the day, Linda gave a presentation at the BISTI meeting. She discribed about 5 projects in which she is involved. It was very interesting, especially to hear about Sal's spherical signatures.

I met with Sal later in the afternoon and made him explain to me all about what he was thinking with regards to the project. Or, at least all that could be reasonably conveyed in about an hour. I found it very helpful. I understand perfectly what a bilinear interpolation is, and should be able to implement it quite speedily. I see now that I didn't need gnuplot, and will probably end up writing to ppm or pgm format, but that is alright.

I have finished the matrices that I will end up using for the length-angle signatures. I need to just write to the ppm files and figure out what the best format is to display in html. I will probably just convert ppm to another format.

We had a farewell lunch today for Tasha at the Still Life Cafe. Another guy working on the UrbanSim project was there too. I forget his name, but he was an undergrad here. It seems there are a number of people on that project.

In the afternoon, I played bridge again. It was quite nice:)