Presentation Slides

https://docs.google.com/present/edit?id=0AbAZnxxQB0bbZGNzY2oyNGJfMTM5aHd2dHFmYzQ&hl=en

1 Pager new link

Hi,

I am sorry I forgot to refresh the link yesterday. here is my one pager.

Meeting 4/26 Notes

It was decided that Patrick and Nate will do the presentation.
Patrick will present Introduction&Background along with experiments: sections 1-3, 8
Nate will cover the game description and belief update methods: sections 4,5,6

We pooled our raw slides. Next meeting is on Wednesday at 2pm. Nate & Patrick will deliver mock presentation.

I'm not entirely sure how the discussion should be presented, or if it should...

Wednesday 4/21 Meeting Notes

Meeting Notes:

We began with some discussions on notations within the paper:
i) theta seemed a bit abused as theta_j is the REAL probability of seller j being honest
ii) The purpose of mu was confusing

We all agreed to read sections 4,5,6 closely to break down any confusion.

After some discussion we divided slide development in the following way:
Joo: Section 5 & 6, Belief update
Guan: Section 8, Experiments & Results
Patrick: Section 1-3, Introduction & Background
Mehmet: Section 4, Bayesian game
Nate: Organization & Discussion

Presenters would be decided at the next meeting on Monday 4/26

Slides for 5&6

This is my slides for chapter 5 and 6. It has 9 slides in it.

https://docs.google.com/present/edit?id=0AUW7NfMTKBufZGQ1NDhqdDJfMTNjZDQzMjRnbQ&hl=en

My 5 page slides

Here is my 5 page slides for section 4.

Some points in section 4

Hi all,

As we talked on wednesday there are some points might be mentioned at discussion section or the assumptions. The first assumtion is that the online nature of the merket is assumed to be eBay-like. The stage games are assumed to be Bayesian games with assymetric information. I do not know if it is enough just to state that assymetric info means a player has more info than the other. One more thing is that perfect bayesian equilibrium thing we discussed on wednesday.
and at the stage game it is assumed that each seller acts as a honest merchant in a fraction of transactions. And at page 6's second paragraph, a term is used like Byeasian decision theory, but this is, I think , like perfact Bayesian Game should be included in a seperate section.

Lastly, there is a subtle point I did not understand. the a and B factors are taken into consideration only when the seller thinks that he is selling a lemon. But, even when the seller thinks that he is delivering the correct item, the buyer still can argue that it is not actually so. I mean when B (beta) is 1, the a (alpha) might still be say 0.5. but there is not such a state in the tree. since the author himself is saying that this a and B are subjective estimate, it is perfectly ok to expect this scenerio I think.

that is all from my section, I know most of them does not worth to be taken in discussion section, but I wanted to post them though.

Next Meeting

Group meeting.
Hey group, how about meeting on Wednesday to discuss how we are going to divide the work in creating slides. How about 1pm? In Dr. Oh's lab again?

my 1 page summary.

1-pager

Guys, here is my 1-pager.  Hard to compress everything I wanted to talk about, but it hits the three major points Prof Oh wanted.  Are we going to turn the meeting notes from last week into the 1-pager for the argument as to why we chose Mehmet's paper?

Meeting Log

Today we met to discuss which paper we would pursue.

In the end everyone agreed that Mehmet's paper on reputation management in an auction game was the best candidate paper the reasons discussed were:

1. Has actual experimental results and comparison
2. Discusses Bayesian and Evolutionary games so it relates to class material.
3. Was long and dense enough to create several slides and require some indepth discussions.
4. Was on the topic of Auctioning, which we have found is a desired skill in a real-world job market.

Furthermore, we eliminated other papers on the following basis:

1. Patrick's social network paper was too short and was on the Shapley value which may not tie that closely into lecture material.
2. Nate's paper was on Satisficing games which also may not tie that well into lecture material, and satisficing games seem a complex subject.
3. Guan's paper, while an application of game theory, was too theoretical with theorems and proofs and no experimental results.
4. Joo's paper the experimental results did not seem as easy to discuss as Mehmet's paper.

We've all agreed to read the Bayesian paper in depth and try to update the blog with questions/comments/etc. based on this thorough reading.

Papers

I skimmed through the paper Guan suggested, it looks very interesting and I would vote for this paper to be the group paper, my only concern is that I think Dr. Oh will say that it is too theoretical. While I agree that it involves game theory being applied to a specific problem, I really think he is looking for something with an algorithm and a measurable a result (graphs, charts, comparison with other algorithms). The results in this paper are theoretical not really empirical.

Guan if you have anything in the same vein of Network Science and Game theory which is a bit more application/empirically based I'd be interested in that paper for the group.

After carefully reading the paper I suggested I am also not a huge fan, it is riddled with poor explanations and some grammatical errors. I'm going to look one more time for papers and will post before the group meets today.

"Bayesian game theory in practice: A framework for online reputation systems"

Guys,

Here is another paper related to online auction sites, specifically eBay. Ill write my summary on this.
http://www.cs.helsinki.fi/u/ptnurmi/papers/nurmi_bayesian_games_reputation.pdf

Mehmet

Meet

Hi all,
Since we have to meet once, how about during class time on Thursday since everyone should be free. I'd also like to suggest a paper on social networks: www.ifaamas.org/Proceedings/aamas08/.../pdf/.../AAMAS08_0345.pdf

which uses the shapley value.