http://portal.acm.org/citation.cfm?doid=1880022.1880029
Basically, the authors discovered a side channel attack on encrypted VoIP systems that use VBR encoding (this is relatively common), and they use HMMs to implement their attack.
http://portal.acm.org/citation.cfm?doid=1880022.1880029
Basically, the authors discovered a side channel attack on encrypted VoIP systems that use VBR encoding (this is relatively common), and they use HMMs to implement their attack.
In Mar 22nd’s class we are going to talk a little more on Adaboost, SVM or other related machine learning techniques. Ben will give you an introduction and I will talk more on applying machine learning techniques to build poker bots. The algorithms we are going to talk about are: SVM, CBR and Bayesian network. They can be downloaded at here.
For the final project, we will target one (or more) of the games used in the 2011 Computer Poker Competition. These games are all Texas Hold ‘Em games, with three structures:
We can decide to either agree to have all teams target one of these games (which will enable a full class competition, as well as more sharing of development), or to have teams target different games.
If you have thoughts on this, please post them here or let me know. In particular, note if you have a preference for which of the three games to target.
I’m uploading slides that were used in the Mar 3rd presentation.
I’ll also add some comments to discuss issues in them.
*pdf: poker
On March 17th, I will be presenting two machine learning algorithms, Support Vector Machines and Adaptive Boosting. I recommend reading the following papers:
A version of our bot capable of beating the benchmark is available below. It was compiled and tested on 32-bit Ubuntu.
No warranty on keeping the dealer/your bot/your computer working.
Please use caution.
Click here to download this “cheater” bot.
Me and Yan Huang will be leading the class on March 15th on Counterfactual Regret Minimization. Seems to be a technique that many of the better bots have been exploring in recent contests. Reading materials include the Rubin and Watson survey paper, the paper from UAlberta that introduced this technique to poker, and Chapter 4 from this textbook Algorithmic Game Theory
Remember that the preliminary deadline for Project 1 is today. The preferred way to satisfy this is to post your bot on the course site, which two teams have already done. If you prefer not to release a preliminary bot, or are making progress on a strategy that has not produced a benchmark-beating bot yet, you can also satisfy it by sending me an email.
Note that there is a bug in the isValidAction in game.c that means the dealer will allow invalid raises. The fix is below. See the comments here for a more complete discussion.
Replace line 822 of game.c with
return 0; // was -1
(Boolean types and strict type checking are good things.)
©2011 cs6501: PoKER | Powered by WordPress with Easel | Subscribe: RSS | Back to Top ↑
Recent Comments