Lab
Time Line for Lab 9 (Decision Tree Learning) and Lab 10 (AI and Humor)
13/11/12 07:56 Filed in: 480
Since we’re a little behind with the material, Lab 9 on Decision Tree Learning is pushed back one week, with the deadline of Tue, Nov. 27. If you want to, you can get a head start on Lab 10, which is independent of the topics discussed in class.
0 Comments
Lab 7 Submission Mode
30/10/12 10:51 Filed in: 480
I’ve modified the Lab 7 entry in PolyLearn to allow electronic submission. Otherwise drop off a printout in the lab. As usual, there's also a Web form associated with it.
Lab 4 Grading
22/10/12 22:04 Filed in: 480
I’ve just completed the grading of the L4 submissions so far. This is one of my favorite labs to grade. It’s my impression that many of the students really get into this, and use the opportunity to apply issues relevant to AI to those entertainment characters. This makes it easy for me to give almost everybody a good grade ;-)
Of course I’m also learning a lot about movies, TV series, and games that I wasn’t aware of. There are the usual suspects that show up a lot (HAL, Sonny, Terminator, Davids from two movies, “Artificial Intelligence” and “Prometheus”, C-3PO and R2-D2, Bender, GladOS, Matrix), but even for those, there’s usually something interesting in the answers.
Right now, I'm essentially the only one getting to enjoy all of the interesting write-ups. I’ve thought about turning this into a more interactive and collaborative version, but didn’t get far enough to actually put it into practice. PolyLearn probably can do a much better job here than Blackboard, but I think it's still a bit limited. I tried out the Semantic Media Wiki that I'm using for the AI Nuggets, but it quickly became too involved for me to do it all by myself.
Here's what I tried to set up: A repository of such characters, with basic descriptions, accompanied buy potential criteria for evaluation purposes. Each student has to submit something that involves at least the following:
If anybody’s looking for an independent study or senior project idea, this might be an interesting one. Since there is some overlap with the chat bot lab, it might also be possible to combine the two, or at least use the same infrastructure. Another option is to use an existing system like Quora or Reddit, and make this available on the Web. This, of course, might be a nightmare from a grading perspective …
Of course I’m also learning a lot about movies, TV series, and games that I wasn’t aware of. There are the usual suspects that show up a lot (HAL, Sonny, Terminator, Davids from two movies, “Artificial Intelligence” and “Prometheus”, C-3PO and R2-D2, Bender, GladOS, Matrix), but even for those, there’s usually something interesting in the answers.
Right now, I'm essentially the only one getting to enjoy all of the interesting write-ups. I’ve thought about turning this into a more interactive and collaborative version, but didn’t get far enough to actually put it into practice. PolyLearn probably can do a much better job here than Blackboard, but I think it's still a bit limited. I tried out the Semantic Media Wiki that I'm using for the AI Nuggets, but it quickly became too involved for me to do it all by myself.
Here's what I tried to set up: A repository of such characters, with basic descriptions, accompanied buy potential criteria for evaluation purposes. Each student has to submit something that involves at least the following:
- one or more characters
- mini-discussions for three or more criteria and how they relate to the selected character(s)
- two or more responses to previous discussions
If anybody’s looking for an independent study or senior project idea, this might be an interesting one. Since there is some overlap with the chat bot lab, it might also be possible to combine the two, or at least use the same infrastructure. Another option is to use an existing system like Quora or Reddit, and make this available on the Web. This, of course, might be a nightmare from a grading perspective …
Lab 3 BotEnvironment Questions
07/10/12 17:16 Filed in: 480
Please post questions about the BotEnvironment on the PolyLearn forum “BotEnvironment Issues”; it’s under “2 - Intelligent Agents” on the course overview page on PolyLearn.
This has several advantages over just sending it to me via email:
- somebody else may have similar issues, and possibly a solution
- you may get an answer faster than from me
- you may get a better answer
- there is a team working on a revision or re-implementation of the environment, and it’s helpful for them to know where the typical problems are
- I’m counting this as evidence towards class participation
I have the subscription mode set to “auto subscription”, which means that everybody in class who doesn’t opt out received a notification of new postings. So posting it there will reach me just as fast as a direct email.
This has several advantages over just sending it to me via email:
- somebody else may have similar issues, and possibly a solution
- you may get an answer faster than from me
- you may get a better answer
- there is a team working on a revision or re-implementation of the environment, and it’s helpful for them to know where the typical problems are
- I’m counting this as evidence towards class participation
I have the subscription mode set to “auto subscription”, which means that everybody in class who doesn’t opt out received a notification of new postings. So posting it there will reach me just as fast as a direct email.
Lab 3 Available
02/10/12 11:07 Filed in: 480
Lab 2 Grading Process
02/10/12 11:03 Filed in: 480
The way I’ve set up the grades on PolyLearn requires me to enter a score for all components of Lab 2, including the answers students submitted via a Webform. For this and a few other labs and assignments, I’ll do much of the grading via demos during the lab. To record the scores for the demos, with my current setup I also need to submit scores for the Web form and the README.txt file. In situations where I can’t do this right away, I’ll enter a score of 0 for those components for now, which I’ll change when I grade the Web form submissions.
Lab 2 PolyLearn File Limit Extended
30/09/12 11:46 Filed in: 480
While I selected “multiple files” for submissions of Lab 2 via PolyLearn, I didn’t realize that the default was set to 3; I extended it to 20. You can also put multiple files into a .zip archive.
Lab 1 Deadline Extended to Thu, Sep. 27
26/09/12 14:20 Filed in: 480
Since I announced the first lab in at least one of the two sections last week on Thursday only, I'm extending the deadline until Thursday, end of the day (23:59).
480 Article Review Opportunity: "Chaotic exploration and learning of locomotion behaviors" Shim Y., Husbands P., in Neural Computation 24(8): 2185-2222,2012
24/09/12 21:18 Filed in: 480
I have received a request to review an article for the ACM Computing Reviews. It's of moderate interest for me, but if anybody from this class is interested, let me know. This could be the basis for your AI nugget presentation, or use it as extra credit, or for replacing a lab, (partial) assignment or quiz.
F.K.
===================================================================
Review of an already published article for ACM Computing Reviews and ComputingReviews.com (http://www.computingreviews.com)
(This review will be published on the ComputingReviews.com Web site, and is not about acceptance or rejection. It is intended to provide feedback on already published work. )
"Chaotic exploration and learning of locomotion behaviors" Shim Y., Husbands P.
Neural Computation 24(8): 2185-2222,2012
CCS Classification:
Abstract:
We present a general and fully dynamic neural system, which exploits intrinsic chaotic dynamics, for the real-time goal-directed exploration and learning of the possible locomotion patterns of an articulated robot of an arbitrary morphology in an unknown environment. The controller is modeled as a network of neural oscillators that are initially coupled only through physical embodiment, and goal-directed exploration of coordinated motor patterns is achieved by chaotic search using adaptive bifurcation. The phase space of the indirectly coupled neural-body-environment system contains multiple transient or permanent self-organized dynamics, each of which is a candidate for a locomotion behavior. The adaptive bifurcation enables the system orbit to wander through various phase-coordinated states, using its intrinsic chaotic dynamics as a driving force, and stabilizes on to one of the states matching the given goal criteria. In order to improve the sustainability of useful transient patterns, sensory homeostasis has been introduced, which results in an increased diversity of motor outputs, thus achieving multiscale exploration. A rhythmic pattern discovered by this process is memorized and sustained by changing the wiring between initially disconnected oscillators using an adaptive synchronization method. Our results show that the novel neurorobotic system is able to create and learn multiple locomotion behaviors for a wide range of body configurations and physical environments and can readapt in realtime after sustaining damage.
===================================================
Article Review
The article has already been published, and the review is intended for additional feedback; if accepted, the review will appear in the ACM Computing Reviews (deadline is probably around Oct. 18).
My expectations for a review are:
- you read the article (which I will send you); usually I read such articles at least twice
- you will check the Internet for related articles (the ones mentioned in the bibliography section, and possibly others the authors may have omitted); typically I won't read all of those, but I'll look at the abstracts, skim the most relevant ones, and read those that help me understand the one under review
- use your best judgement to evaluate the quality of the article, based on a review form provided by the journal editors
- write a review for people interested in the topic; this one is usually short (about 250 words)
Overall, such a review typically takes me 5-10 hours; count on spending maybe a little more if you're not familiar with the topic. In both cases, I will act as the main reviewer, and reserve the right to use or omit your contributions. I will acknowledge your assistance to the editors. If I use any of your contributions for the ACM Computing Reviews, your name will also be forwarded as a co-author of the review. In that case, you need to sign a copyright release form.
F.K.
===================================================================
Review of an already published article for ACM Computing Reviews and ComputingReviews.com (http://www.computingreviews.com)
(This review will be published on the ComputingReviews.com Web site, and is not about acceptance or rejection. It is intended to provide feedback on already published work. )
"Chaotic exploration and learning of locomotion behaviors" Shim Y., Husbands P.
Neural Computation 24(8): 2185-2222,2012
CCS Classification:
Abstract:
===================================================
Article Review
The article has already been published, and the review is intended for additional feedback; if accepted, the review will appear in the ACM Computing Reviews (deadline is probably around Oct. 18).
My expectations for a review are:
- you read the article (which I will send you); usually I read such articles at least twice
- you will check the Internet for related articles (the ones mentioned in the bibliography section, and possibly others the authors may have omitted); typically I won't read all of those, but I'll look at the abstracts, skim the most relevant ones, and read those that help me understand the one under review
- use your best judgement to evaluate the quality of the article, based on a review form provided by the journal editors
- write a review for people interested in the topic; this one is usually short (about 250 words)
Overall, such a review typically takes me 5-10 hours; count on spending maybe a little more if you're not familiar with the topic. In both cases, I will act as the main reviewer, and reserve the right to use or omit your contributions. I will acknowledge your assistance to the editors. If I use any of your contributions for the ACM Computing Reviews, your name will also be forwarded as a co-author of the review. In that case, you need to sign a copyright release form.
Lab 1 Available
19/09/12 16:52 Filed in: 480
The first lab activity is available now. You can choose between three alternatives: An evaluation of chatbots, an exploration of Hunch.com, or an examination of educational modules related to AI.
The first one on chatbots is the most developed one, the other two are somewhat experimental.
Each alternative is worth 10 points for a total of 100 in the labs.
The deadline is Tue of Week 2, end of lab period.
The first one on chatbots is the most developed one, the other two are somewhat experimental.
Each alternative is worth 10 points for a total of 100 in the labs.
The deadline is Tue of Week 2, end of lab period.
Various Feedback Forms Available
28/11/12 22:07 Filed in: 480
I’m using a few Web questionnaires to collect feedback about various aspects of the class; please check the Class Feedback page for the links.
Lab 2: Auto-Step Solution for BotEnvironment
26/09/12 12:25 Filed in: 480
Here’s a solution for the missing Auto-Step behavior in the environment for Lab2; thanks, Alan!
Author: Alan Phu Truong.
Solution:
Instead of using the following to start your environment in Linux/Mac:
java -classpath .:WumpusEnvironment.jar BotEnvironment.WumpusEnvironment
Use:
java -Xint -classpath .:WumpusEnvironment.jar BotEnvironment.WumpusEnvironment
On Windows, replace the colon with a semicolon:
java -Xint -classpath .;WumpusEnvironment.jar BotEnvironment.WumpusEnvironment
AutoStep seems to work under more circumstances with this flag.
I’ve also added a pointer to additional material on Blackboard (under BotEnvironment JavaDocs, BotEnvironment Tutorial, and BotEnvironment FAQ) to the Lab 2 description.
If you want to be updated about postings to the Blackboard Discussion forum, click on the forum title, and then on the “Subscribe” button at the top. Subscriptions can be enabled or disabled for a specific forum, so not all of them will have the “Subscribe” button.
Author: Alan Phu Truong.
Solution:
Instead of using the following to start your environment in Linux/Mac:
java -classpath .:WumpusEnvironment.jar BotEnvironment.WumpusEnvironment
Use:
java -Xint -classpath .:WumpusEnvironment.jar BotEnvironment.WumpusEnvironment
On Windows, replace the colon with a semicolon:
java -Xint -classpath .;WumpusEnvironment.jar BotEnvironment.WumpusEnvironment
AutoStep seems to work under more circumstances with this flag.
-Xint: Operate in interpreted-only mode. Compilation to native code is disabled, and all bytecodes are executed by the interpreter. The performance benefits offered by the Java HotSpot VMs' adaptive compiler will not be present in this mode.See http://download.oracle.com/javase/1.3/docs/tooldocs/linux/java.html
I’ve also added a pointer to additional material on Blackboard (under BotEnvironment JavaDocs, BotEnvironment Tutorial, and BotEnvironment FAQ) to the Lab 2 description.
If you want to be updated about postings to the Blackboard Discussion forum, click on the forum title, and then on the “Subscribe” button at the top. Subscriptions can be enabled or disabled for a specific forum, so not all of them will have the “Subscribe” button.