Recently came to know about SIGBOVIK. Unluckily too late for this year. 

This entry is an excuse for the lack of a good collaborative site. 

How to Select Courses
I consulted with my supervisor about the courses. I am in ECE and my department allowed me to take any number of courses from other faculties. I took several courses (see my courses here) from SoC (School of Computing) as they were more relevant to me. Modules from ECE, Maths, Comp, ISE and Stat. Other courses can be found from the list of all faculties.

Typically you get a week at the start of the semester to decide on your courses. In the first week you can visit a course, talk with the lecturer and by the end of week can decide on the course. During the first week, there is also a second-hand book buying/selling program (library has a huge collection of books, but usually there is only 1-2 copies of a book). You can also use the "Used Textbook Forum" in Ivle Student Workspace, but this forum is dominated by undergrads (as many other things in NUS). 

Seminars and Talks
I was interested in AI Planning-related SoC talks. I found out that there is a mailing list of COMP seminars. I subscribed to it at seminar-l at comp.nus.edu.sg (you can subscribe by sending a mail to 'seminar-l-request@comp.nus.edu.sg' with subject 'subscribe'). I can submit my attendance to these seminars as well as part of my module requirement on seminars in my department. Similarly I subscribed to mailing lists of other faculties (sent them an email) I was interested in.

Conferences and Workshops
After, say 6 months-1 year, your advisor may ask you to prepare for a workshop/conference. Searching for good conferences/workshops in your area may be tricky at the start. There are some conference lists like allconferences.com, but usually you have to look at more than one sites to find good conferences of your choice. Some profs maintain conf list on their websites, which is quite useful. Also beware of some conferences of poor quality: they have a wide subject cover (little focus) and will publish all papers they receive.

Searching
Search plays an important role in research. Apart from Google, I use Google Scholar. As a NUS student, I can access IEEE, Springer etc via the library gateway. To keep up to date with the current research, it is useful to subscribe to mailing lists in your area - ask your supervisor if there are any. As for me, I am subscribed to:

http://groups.google.com/group/search-list?hl=en
http://www.euro-online.org/eume/
http://www.jair.org/

among others.

Housing/Accommodation
A guide for graduate students in Singapore would be incomplete without the topic of housing. Housing is tricky in Singapore, so read on.

Most houses available for rent in Singapore have 3-4 rooms. Therefore, usually singles (and even married couples) share a house. The common terminology is "Master room" means a room with attached toilet and bathroom and a "Common Room" has a shared toilet and bathroom. Cost for a master room would be 700-1200 SGD per month, common room would be lesser. Separate houses (1+1 -- the trend/law is that the house owner locks one or more rooms of the house, for example lock one room of 2+1 will leave 1+1 available to rent) are 0.9K-1.5K or even more, depending on location.

A map of NUS is quite useful at the start (see for example the campusmap). The close-by areas to NUS are Clementi, Dover, Commonwealth, Jurong East, Jurong West (also called Boon Lay) and Pandan Gardens. There are usually direct buses from these places to some parts of NUS (there are many routes on the various sides of NUS -- Pasir Panjang Road, Clementi Road, AYE -- see this link). You will need Ezlink card to travel on these buses (you can get Ezlink card on any MRT station -- Ezlink card is also used within MRT). Within NUS you can travel using the Internal Shuttle Buses (ISB) named A1, A2, B, C, D -- these are free buses.

Finding accommodation can be a little tricky with agents involved. The agents ask for a commission of 1 month rent for up to 1 yr contract and 2 month rent for a 2 yr contract. Be careful that you trust the house owner (check with other tenants for example), otherwise there are cases where owners are not very friendly.

A good site to find accommodation is sg-house.com. There is also an internal NUS site for accommodation for NUS students. This internal site has houses close to NUS. Further, there are usually some pamphlets about rent at bus stops near Central Library and opposite to it.

I welcome your comments.

Also see my article on: Productivity Tools

In our lab, we are building a real-time planning system in C++, called Crackpot. A planning system can reason about which set of actions need to be taken in order to achieve specified goals. For example, if a robot Wall-e is asked to fetch a drink from the local store, it must internally generate a plan to reach the store, buy the drink and then return back home.

In our related games lab, Crackpot will be used in a story planning project. It will serve as an engine to automatically develop plans or "recipes" for story content generation, with the goal to optimize player experience. Crackpot is a general (the technical term is domain-independent), real-time, dynamic planning system. Being general, it can be used for automated planning across diverse applications (not just gaming).

Crackpot is a public domain, open source software. See https://sourceforge.net/projects/crackpot/. See also my blog on Crackpot.

Recently there has been an interest in our lab about exploring the possibility of using our planning system in the security domain. While the current subject is Honeypots, I spent a little bit of time exploring other existing usages of AI planning in security. I briefly saw a paper(pdf) on "Identifying Terrorist Activity with AI Plan Recognition Technology" but that is not the topic of this post.

I read up in more detail about the CyberSecurity domain that appeared in ICAPS 2009 IPC.

The main paper that discusses the CyberSecurity domain and planning is:

Course of action generation for cyber security using classical planning (2005)
by Mark Boddy ,  Johnathan Gohde ,  Tom Haigh ,  Steven Harp
In Proc. ICAPS 2005
Download: http://www.adventiumlabs.org/files/ICAPS05BoddyM_0.pdf
Abstract: We report on the results of applying classical planning techniques to the problem of analyzing computer network vulnerabilities.  Specifically, we are concerned with the generation of Adversary Courses of Action, which are extended sequences of exploits leading from some initial state to an attacker âs goal. In this application, we have demonstrated the generation of attack plans for a simple but realistic web-based document control system, with excellent performance compared to the prevailing state of the art in this area.  In addition to the new capabilities gained in the area of vulnerability analysis, this implementation provided some insights into performance and modeling issues for classical planning systems, both specifically with regard to METRIC-FF and other forward heuristic planners, and more generally for classical planning. To facilitate additional work in this area, the domain model on which this work was done will be made freely available. See the paperâs Conclusion for details.

The work was motivated by the problem of analyzing the threat posed by malicious insiders. The Behavioral Adversary Modeling System (BAMS) prototype generates hypothetical attacks against a simple but realistic model of a web-based Document Management System. BAMS can be used to generate a sequence of attacks against a given network, exploiting different vulnerabilities as the old ones are fixed by configuration changes. Hoffman's METRIC-FF planner was used for generating the Course of Actions.

Given this Behavioral Adversary Model, analysts can make intelligent trades among different security architectures and counter-measures. An example goal for our canonical bad guy Bob may be to minimize the detection risk of the plan while gaining access to a secret document:
(:goal (knows bob secret_info))
(:metric minimize (detection_risk))

Some critical comments from me: The implementation was able to show near real-time performance numbers on a moderately-sized domain in the addressed problem. The paper has a good overview on the "Other Approaches on Vulnerability Analysis". The domain is quite a large description -- its 30 problems in ICAPS 2008 is 600MB of text. A cursory look at the domain reveals that the main reason for this is the redundancy required by the similar action specifications in the domain. An alternative, possibly hierarchical modeling of actions may be useful in reducing the size of the domain description and making the description more natural for humans.

Here is a true story on cattywumpus.

I will post all new research-related content here. So there will be no more updates on my previous research page at http://palgos.blogspot.com/ I am also likely to move the most of the content from my previous research page to this blog.