Countdown to Jen's Dissertation Defense: April 1, 2005

Monday, January 31, 2005

61 days left

Chapter 6: Continuous networks

I figured out how to present an analysis for path length. The results are beautiful. The code is running now to get results for paths of length 3 and 4. Those should be done in the morning and I'll write up this section tomorrow. Tomorrow should also see the start of the algorithm section (at least in text) and perhaps the implementation to get some results.

Sunday, January 30, 2005

62 days left

Chapter 2: Social Networks
Submitted this chapter as a journal paper to Social Networks Journal

Chapter 6: Continuous networks

The never ending chaaaaaaaaa aaapterrrrrr.....da da da...da da da...da da da

I finished the justification for why the small world properties prevent an accurate analysis of path length with the common neighbor accuracy approach that I used for the section about the effect of trust.

Now, I need to come up with some way to discuss the impact of path length. I think I've figured this out using distributions of expected values with the probabilities discovered in the trust section. I'll hopefully get that working tomorrow. If it works, I'll be finished with this section and can go on to the algorithms part.

Saturday, January 29, 2005

63 days left

Yet another unproductive day today. There are just other things I have to get done like bathing, the most rudimentary house cleaning (i.e. after 4 days, the dishes REALLY shouldn't be in the sink), and today dealing with a mouse in the house and a leaking faucet.

Chapter 6: Continuous Networks

I ran a few more sets of experiments and put together the (finally) correct data for length 3 paths. However, it was not coherent with averages that jump around and wildly varying numbers of data points (sometimes only 4 and other times over 400 with no correlation to the parameters). It seems that this is probably due to the connectance because of the small world network. So I've started writing up this description for the path length section. About 337 words so far. I'll have issue pretty well justified but eventually I'll still have to come up with something to suggest how path length will affect accuracy (or a good reason to ignore it). I could potentially justify this with the definition of trust similar to the transitivity argument from chapter 2. This is something I'll have to think more about.

Chapter 4: Background

Read a couple of important papers on this list. They didn't have all that much relevant stuff.

Friday, January 28, 2005

64 days left

Unfortunately, I didn't get much done today because I had another paper to work on. I did manage to run some experiments for chapter 6 on accuracy over connections with 2 intermediate links. There are some duplicates in there messing things up, but the results aren't all that coherent yet. More work on this tomorrow.

Thursday, January 27, 2005

65 days left

Chapter 6: Continuous Trust

Wrote >760 words in this chapter, adding in the information about the trust distributions and completing that section of the paper. The total is now at 11 pages and 2356 words.

Wednesday, January 26, 2005

66 days left

Chapter 6: Continuous Networks

I ran the ANOVA and correlation tests for the social network experiments and randomized data. Wrote just over 1100 words in the chapter which completed the introduction to section 2 (Patterns of Trust) - describing the changes that needed to be made from the binary algorithm - and section 2.1 - the correlation of trust and accuracy. This chapter is now 8 pages. I have all the material I need for the next section that discusses distribution of ratings, and will hopefully complete that tomorrow.

Tuesday, January 25, 2005

67 days left

Chapter 6: Continuous networks

Today, I set up and (re)conducted the first set of experiments that will form the foundation of this chapter. I took each node, looked at each of its neighbors, and calculated the difference between the node's rating and the neighbor's rating of common neighbors. Each data point contained the rating of the neighbor and the difference. I sorted by the rating of the neighbor node to see the correlation between trust rating and differnece. The results there look good with a nice linear increase in accuracy. I randomized the ratings and did the same thing and it shows no correlation. This is also good.

I need to think about how this can be carried out more steps. There needs to be a way to check this over distance, as well.

Chapter 3: Trust
Submitted this to IJCAI

Monday, January 24, 2005

Current Status

  1. Introduction - nothing
  2. Social Networks - submitted to Jim
  3. Trust - submitted to Jim
  4. Inferring trust - 4 pages(1780 words) complete. The section that generally describes calculating trust is complete. Still working on the background. It needs more to fill out the P2P section, and I think quite a bit more in the social section
  5. Binary trust - submitted to jim
  6. Continuous Trust - 3 pages (460 words) complete. The section that describes the two networks is finished
  7. FilmTrust - 4 pages (748 words) complete. Description of the site is finished.
  8. Other Applications - 3 pages (1309 words) complete. Description of the system is complete. Need to figure out what to present as results for this.
  9. Conclusion - nothing
  10. Future work - 2 pages (489 words) complete. Description of MOAM is complete. Description of more systems is necessary.

Goals for this week
  • Finish chapter 4
  • Get structural results for Chapter 6. Write them up.

68 days left

Aside from the dissertation, I have hard deadlines on a conference paper, a journal paper, and a book chapter. I have soft deadlines on a second journal paper and a code implementation. Today, I had to work on some of these things. I finished up the journal paper (I hope) and basically finished up the conference paper. My book chapter co-author has done his part, which is great. I have my part in perspective, so I'm feeling pretty capable of getting that done on time. As such, there was not a tremendous amount of progress chapter wise.

I am converting the trust definition chapter to a paper for IJCAI.
Chapter 10: Future Work

This chapter didn't exist before today. I wrote about 500 words which completed the section on MOAM.

Sunday, January 23, 2005

69 Days Left

Chapter 6: Continuous Trust

Started this chapter, wrote about 500 words with introduction and background on the two networks

Chapter 7: FilmTrust

More work on code and experiments. I've decided to look only at the AFI films because that avoids situations where, say 2 people rate a movie, and the recommendation to one ends up being the rating of the other person.

Saturday, January 22, 2005

70 days left

Chapter 3: Trust
Completed first draft (addition of ~500 words)

Chapter 7: FilmTrust
More work on code and analysis.

Friday, January 21, 2005

71 days left

Chapter 2: Social Nets
Incorporated editorial changes recommended by Yarden and sent the draft off to Jim.

Chapter 3: Trust

~1000 words in trust chapter. I extended the section on transitivity, found a large set of references to include, and also added a table and text about the values of trust to that section.

Chapter 7: FilmTrust

More work on the code experiments and some preliminary result analysis.

Thursday, January 20, 2005

72 days left

Today's Work

7. Laid out experiments and code for chapter 7 (FilmTrust).
8. Moved over text from TrustMail paper at CEAS. Chapter 8 is now 3 pages.

Overall, this was not a very productive day. With class and a headache and some running around. Hopefully tomorrow will see more work done.

Wednesday, January 19, 2005

73 days left

Current Status

For background, the dissertation contains 10 chapters:

1. Introduction
2. Social networks
3. Trust
4. Inferring trust in networks: background and related work
5. Calculating trust in binary networks
6. Calculating trust in continuous networks
7. Trust Networks in application: Film Trust
8. Trust Mail (may become "additional applications")
9. Conclusions
10. Future Work

Current status on the chapters:

  1. nothing
  2. basically done - needs edit cycle with advisor
  3. 4 pages written with material in 2 of the 3 main sections. needs references
  4. finished. needs only surface proof from advisor. sent to him today.
  5. nothing. this will largely depend on results i get from analyzing the social networks. code for that is written on a surface level. this is arguably the most important chapter and does not have much finished yet.
  6. 5 pages finished, 1 section completely done. related work needs to be done. experiments for 3rd section need to be laid out including questionnaires for users.
  7. nothing formally composed, though much is written up in previous publications.
  8. nothing
  9. nothing

Today's work

~1000 words in Chapter 4.