Artificial Intelligence Research

Workshop on Machine Learning, C-DAC Mumbai
The Knowledge Based Computer Systems (KBCS) division of C-DAC Mumbai is organizing a 2 Day Workshop on Machine Learning at the Navi Mumbai (Kharghar), campus on 19th and 20th December, 2008. Organized at India, the workshop is meant to provide a comprehensive introduction to Machine Learning, focusing on conceptual understanding of popular ML algorithms and practical applications. Apart from covering popular ML techniques such as Artificial Neural Network (ANN), Support Vector Machine (SVM) and Clustering, discussions for modeling a problem for using machine learning including input-output transformation will be carried out. Participants will get hands on experience with these algorithms; using toolkits such as Weka.

The brochure gives you a brief information on the workshop, which can be located at the C-DAC Mumbai website, at this URL. Kindly forward the details about the workshop to all those whom you feel may be interested in participating in the workshop. You may also go ahead and blog about it, post it to mailing lists, groups, and communities etc. Do keep us informed about it.

Contact:
Dr. Sasikumar </a></b></a>

the_little_sasi
Centre for Development of Advanced Computing (Erstwhile NCST), Mumbai
Knowledge Based Computer Systems (KBCS)
Raintree Marg, Near Bharati Vidyapeeth, Sec 7, CBD Belapur, Opp. Kharghar Railway Station
Navi Mumbai, 400614, INDIA
Voice: 91-2227565303, Fax: 91-2227560004
E-mail: kbcs [at] cdacmumbai [dot] in
URL: www.cdacmumbai.in/index.php/cdacmumbai/research_and_publications/research_groups/kbcs_artificial_intelligence/events

Ensemble methods
Ensemble methods, especially bagging and boosting, are well-established methods (for an introduction see here.) But papers on it gives the impression that atleast bagging (using several predictors or classifiers in parallell and e.g. taking their average) is not completely understood.

Most papers agree that variance is reduced given that the classifiers have a sufficient "diversity", but how this ties in with the bias-variance theorem or even mean square error, is unclear. The paper "The Effect of Bagging on Variance, Bias, and Mean Squared Error" by Andreas Buja and Werner Stuetzle offer some leads, but I find no definitive account of these issues.

Do rigorous results on bagging vs bias, variance and MSE exist or is it mostly empirically based?

Where's the best?
If you could go anywhere to do research in AI, where would you go? Who's the leading experts in your subfield, where's the best research groups?

I'm doing an AI thesis and it wasn't until halfway in my thesis that I could have even tried to answear this. The best ones must be there, but they don't seem to leap out at you when you read journals.

Workshop on Rule Based Expert Systems, C-DAC Mumbai
The Knowledge Based Computer Systems (KBCS) division of C-DAC Mumbai, India is organizing a two-day Workshop on Rule Based Expert Systems at the Navi Mumbai (Kharghar) campus on 17th and 18th October, 2008. The workshop is meant to provide a comprehensive introduction to Expert Systems, focusing on the practical application. This workshop is targeted at academicians, IT Managers, consultants, domain experts, professionals working on advisory systems and potentially anyone who feels a need for building systems with human expertise.

The brochure gives you a brief information on the workshop, which can be located at the C-DAC Mumbai website, at this URL. Kindly forward the details about the workshop to all those whom you feel may be interested in participating in the workshop. You may also go ahead and blog about it, post it to mailing lists, groups, and communities etc. Do keep us informed about it.

Contact:
Dr. Sasikumar

strategic games and real economy
Hello. Sorry for my English

My friends and me are interested in everything about cybernetics in real
economy.
May be you heard about Cybersyn -
http://en.wikipedia.org/wiki/Project_Cybersyn. Stafford Beer -
http://en.wikipedia.org/wiki/Stafford_Beer and his
http://en.wikipedia.org/wiki/Viable_System_Model

Cybersyn uses cybernetic algorithms to compress dataflows, show them on
displays in simple graphic way and to predict evolving of Chily economy.
But all solutions were taken by people-operators in control rooms.

We want to stay a questions like
1) what informations do you have about systems like Cybersyn?
2) May be you heard about systems which control country economy fully
automatically.
3) Do you know science works about automatic analyzing and controlling
of country or big enterprises (communities) economies?
4) What opensource strategic games economy engines do you know?
5) Do you know strategic games developers and architectors whom we can ask our questions about cybernetics in real ecnomy?

Thank you for answers!

Drukker chat bot---natural language understanding
I am currently programming an intelligent chat bot based on RelEx, CMU/AbyWord link-grammar, GATE, opennlp, Cyc systems. Right now, it has a very broad coverage of English (link parser's one) and a small amount of concept and question understanding. At the moment, it understands exactly one type of questions: what/who is an (object)? All semantic graphs of sentences are added to a Cyc KB, though currently bot is unable to work with the information it stores in Cyc.

Bot exists in 4 chat networks, I'd be glad if you visit it to chat a bit, right now he is being visited by some of my friends only.

Right now, I plan to work with the model level. That is, how to 1) convert a text into a model of situation, 2) change a world model, and 3) compress a world model into a response text. Namely, I plan to do items 1 and 2 on the example situation about the dog who catches the frizbee thrown by a human in a park, with miscelanneous conditions, reasoning, and events.

People on the #ai/freenode channel say that the bot's NLP is impressive. It is so thanks to the work of the authors of the above-mentioned systems, I work on the bot during just about a week.

Come visit the bot, I'd be glad. :) He is here: icq 380651255, gtalk drukkerbot, irc Drukker on channel #Drukker (on freenode and rusnet).

***

Linguistic features:
- semantic parsing (into a semantic graph);
- resolution of pronouns (not very efficient at the moment).

More details are at http://s6.livejournal.com/592394.html (Russian) and in my livejournal

s6 (Russian).

Comments are very welcome.

***

( An example of semantic and syntax parse )

Reccurent Neural networks and AR processes
I am working with a recurrent neural network with one input and one output, faced with the task of predicting a relatively simple process. I am using linear nodes because this works well and because the sigmoid (tanh) nodes tended to just max out.

So far all is well, but then in (Ref 1) it is claimed that a linear recurrent network will only realize processes similar to the AR process. The reference says "AR-like" but are we talking equivalence here?

Formally: For a given trained recurrent linear network and a bounded interval for the input, do coefficients exist for an AR process so that the AR process will always give the same output as the network?

Ref 1: Page 22 of H. Jaeger, The echo state approach to training an analysing recurrent neural networks. (http://www.faculty.iu-bremen.de/hjaeger/pubs/EchoStatesTechRep.pdf)

Apologise basic Automated Collaborative Filtering Question

I have a question to "describe an approach(es) that can be used to improve the robustness of an ACF* algorithm".

Any help appreciated.

A large AI competition
The Second Annual Reinforcement Learning Competition is under way.
This year's competition include some challenging task including a hovering helicopter, a real-time strategy game, a robocup related task and Tetris. Deadline is july 2008.

Creating a better Go Program (an article by IEEE)
Cracking GO By Feng - Hsiung Hsu
First Published October 2007 < http://www.spectrum.ieee.org/oct07/5552 >

Brute-force computation has eclipsed humans in chess, and it could soon do the same in this ancient Asian game

In 1957, Herbert A. Simon, a pioneer in artificial intelligence and later a Nobel Laureate in economics, predicted that in 10 years a computer would surpass humans in what was then regarded as the premier battleground of wits: the game of chess. Though the project took four times as long as he expected, in 1997 my colleagues and I at IBM fielded a computer called Deep Blue that defeated Garry Kasparov, the highest-rated chess player ever.
You might have thought that we had finally put the question to rest—but no. Many people argued that we had tailored our methods to solve just this one, narrowly defined problem, and that it could never handle the manifold tasks that serve as better touchstones for human intelligence. These critics pointed to weiqi, an ancient Chinese board game, better known in the West by the Japanese name of Go, whose combinatorial complexity was many orders of magnitude greater than that of chess. Noting that the best Go programs could not even handle the typical novice, they predicted that none would ever trouble the very best players.
Ten years later, the best Go programs still can't beat good human players. Nevertheless, I believe that a world-champion-level Go machine can be built within 10 years, based on the same method of intensive analysis—brute force, basically—that Deep Blue employed for chess. I've got more than a small personal stake in this quest. At my lab at Microsoft Research Asia, in Beijing, I am organizing a graduate student project to design the hardware and software elements that will test the ideas outlined here. If they prove out, then the way will be clear for a full-scale project to dethrone the best human players... (full story at the above website)

FSA Induction
Anyone know of a good place to start reading on the state of the art? I have a bunch of references, but they're pretty disjoint or specialized... a (recent) survey paper type of thing would be helpful, as would any recommendations for classic papers in the subject. TIA.

Theory on analytical-AI hybrids
Engineering applications of neural networks sometimes use "hybrid systems" in the sense of combining a neural network with a traditional physical/analytical model. A typical approach is as follows:

Given a set of known input data X and output data Y, we should predict Y from X. Let Y_analytic be the prediction that the oldfashioned model gives. We take the difference between true and predicted: Y_diff = (Y - Y_analytic) and try to train the neural network to predict Y_diff from X. We name the prediction Y_ann

If the neural network does its job, the sum Y_analytic + Y_ann will be closer to the true Y than what the analytic method alone managed. Supposedly, this "hybrid approach" is better than ANN or a physical model on their own.

Seems sensible, but I've only managed to track down the method in papers about engineering applications, with no references to a thorough theoretical discussion of its advantages and drawbacks. Has anyone come across such a discussion?

Artificial Intelligence in Software Development
Hi! I don't remember the last time I saw a post in this community, so I suppose I can be excused for posting something vaguely off-topicish.

I'm trying to make a Java library with AI and heuristics functions. Here's what I'm trying to research:

-Are there many AI algorithms, methods/functions and such that are used commonly in AI programming?

-Are there any common functions used broadly across robotics AI, game AI etc that would be useful to have as a library?

-If there are no universally useful functions to have to call on as a library, are there any functions that could be useful specifically to an area of AI such as games or robotics that I could focus on?

Obviously, all I'm looking for here is throw-out ideas. If you know any similar attempts or related stuff to check out, that'd be nice also. Thanks in advance for comments!

Current Mood: Researching

Embarassing problem
I've got a time-series that looks something like the picture below, but with added noise. I want to use some AI or statistical method to categorize subsets of the time-series according to shapes. Like "box" or "ramp". Not all ramps have the same slope, but if I can detect "slope" and "step change", I can combine that into "ramp".

I just want to get started with some preliminary results and take it from there, so I've been looking for software or simple algorithms that would perform this classification. I've tried a few programs that run in matlab, but they don't work. Can you recommend some program or algorithm that would work well with this problem?

(x-posted to Timeseries)

a neural network problem
I've been trying to make a neural network learn a relatively simple problem, but no luck. So I thought I'd hear if anyone's got any ideas.
( explanation with pictures )

Linguistics v. Cognitive Science

So, I'll be applying for doctoral programs in the fall and I have a question for those who are out there in the industry/academe. I'm working towards a career as a professor (or other research position, perhaps) and am curious about the merits of getting my doctorate in linguistics itself vs in cognitive science with a linguistics focus.

I have a BA in linguistics (with anthropology, and some psychology) and will be finishing up an MS/MSE in computer science (with a focus in artificial intelligence, and language/compiler design) next spring. I'm interested in the whole programme of cog.sci, though I'd like to get a professorship focusing on linguistics (or the moral equivalent for schools that lack a department as such). My interest in CS is the aforementioned AI (particularly biologically inspired, e.g. genetic algorithms, swarm intelligence, neural nets) and language/compiler design, as well as the theory end of systems science. Interest in computational linguistics is more along the lines of natural language processing and other more theoretical linguistic topics, rather than data mining or statistical analyses. Interest in linguistics is split between theory (esp. morphosyntax; OT; agglutinative syntax) and sociolinguistics (esp. re the effects of technology on society/language; gender/sexuality; Japanese and Korean).

My question is, how are cog.sci degrees received by the linguistics community? (or the CS/AI community for that matter?) I figure the discipline's been around long enough that it's not entirely unheard of, but I wonder if it would hamper my career goals by being too focused/outre. I know there's a glut of theoretical linguists out there, so it could also provide a nice edge. Also, I'm curious which cog.sci programs y'all think are good for the linguistic bent?

crossposted: linguistics, linguists, compling, ai_research, neurophilosophy

CSI Special Interest Group on Artificial Intelligence announces a Symposium on AI in Industry
CSI Special Interest Group on Artificial Intelligence announces a Symposium on AI in Industry
AI methodologies have been effectively applied to solve a variety of complex problems such as fault prediction, logistics handling, intelligent machine interfaces, automated machine translation, handling large document collections, and so on. As computer systems move from handling routine problems to higher level problems and tasks, AI will be an important ingredient of tomorrow's software solutions. In an era of increasing competition, industries moving from mass-production to personalization, these technologies play a pivotal role. However, most organizations do feel wary in using AI techniques. It is in this context, that SIGAI is organizing a symposium focused on real-life applications of AI in industry (not restricted to IT), as a part of the forthcoming International Joint Conference on Artificial Intelligence (IJCAI-07).

The symposium will include

# Talks from industry leaders on the role of AI in their fields
# Case studies of deployed AI applications

Date: January 10, 2007
Venue: HICC, Hyderabad

Please visit the website : http://sigai.cdacmumbai.in for more details.

Emotiv Systems Seeking Volunteers for Paid Research

EMOTIV SYSTEMS SEEKING VOLUNTEERS FOR BRAIN RESEARCH

Extra compensation for Emotiv Systems Research Participants!

Our world class team of scientists is currently researching human emotion during
multimedia interaction. We are currently seeking volunteers to join us at
our Pyrmont office to partake in this exciting new research.

Our current experiment involves watching short film clips whilst having an electrical profile (EEG) of your brain taken, to determine the electrical activity of the brain and what parts of your brain are active during different emotional states and mental tasks.

This recording is taken using a 100% safe and painless headcap of sensors in our Pyrmont Wharf office overlooking Sydney Harbour [jonesbaywharf.com.au].

All participants will receive $20 to compensate their time as well as images of themselves in the headcap setup, a comprehensive personality assessment, and an emotional profile detailing their experimental performance.

If you have any questions or wish to arrange a time to come please contact us
by phone [9552-2559] or email;

Deborah Johnston (deborah@emotivsystems.com)
Michael Orr (michael@emotivsystems.com)

Feel free to distribute this email to anyone who you think may be
interested as all are welcome.

Warm regards,
Deborah Johnston

--
Deborah A. Johnston
Research Scientist

Emotiv Systems Pty Ltd
Suite 12, The Upper Deck
Jones Bay Wharf 19-21
26-32 Pirrama Road
Pyrmont NSW 2009
Sydney, Australia

T: +61 2 9552 2559
E: deborah@emotivsystems.com

http://www.emotiv.com

Emotiv Systems December Recruitment Poster

Emotiv Systems December Recruitment Poster

Advertisement