Saturday, April 30, 2005

 

Nietzsche on tape

Just finished a terrific Teaching Company video course Will to Power: The Philosophy of Friedrich Nietzsche taught by husband and wife tag team Kathleen Higgins and Robert Solomon of the University of Texas at Austin. The sequence and choice of material, and the consistent quality of the lectures made this series a joy to listen to. Of course, I've already confessed to being a Teaching Company fan before (see here).

If you exercise on a treadmill or stationary bike, the video courses are a great way to pass the time and improve your mind. If you run and listen to music or tapes, the audio tapes or CDs are the ticket. Numerous courses are available on a variety of subjects, and in a decade or so of using Teaching Company materials, I've never had a clunker course.

 

Blogs: Audience will follow trust

Hugh Hewitt has a longish post on the transformation that is taking place in mass media, partially a result of blogging. His key point is that whereas once trust followed large audiences, today audiences follow trust. Although Hewitt focuses on news and commentary media, the longer term effects on business and academic life could be just as important.

Thursday, April 28, 2005

 

Martin Seligman kick

I'm on a serious Martin E. P. Seligman reading kick. If you haven't done so, take a look at the following books:
Seligman, M. E. P. (1998). Learned optimism. New York: Free Press.

Seligman, M. E. P. (2002). Authentic happiness. New York: Free Press.

Peterson, C., & Seligman, M. E. P. (2004). Character strengths and virtues: A handbook and classification. New York: Oxford University Press.

These are not touchy feely self-improvement books (although, there are some of those kind I recommend as well). They are carefully written books based on a growing body of peer-reviewed research in positive psychology that concentrates more on mental health and a good life than on mental illness and life's problems.

If you are interested in happiness and achievement, read them in the order above, and take all the quizzes and self-assessment tests.

Tuesday, April 26, 2005

 

A P2P Genetic Algorithm Environment for the Internet

I was reading the last issue of the Communications of ACM (April, 2005; pp. 113-116) and found an article on a P2P environment for GA applications.

Pier Luca

 

Special issue on computational intelligence and soft computing of JISLM

Ying-Ping Chen and Jian-Hung Chen (former IlliGAL members) are organizing a special issue for the international journal of information systems for logistics and management (JISLM). JISLM is a quite new journal but GEA friendly. This special issue on computational intelligence and soft computing welcomes any GEC related paper. It's always good to have another GEC friendly journal :-) For more information, please visit here.

Monday, April 25, 2005

 

Toward conscious machinery

Martin Butz had an interesting comment on my post about machine consciousness (it deserved post status), so I'm quoting from it here:

Certainly I agree that also consciousness is an emergent property composed out of many modules, many (local) interactions through connections that are biased in one way or the other. What might be the most important ingredients of consciousness?
He goes on to cite Deacon's The Symbolic Species and the European project MindRACES: From Reactive to Anticipatory Cognitive Embodied Systems. There is lots to like in the discussion:

To sum up, it seems to me that in order to proceed towards machine consciousness it is necessary to have at least (1) a body that can manipulate a world and (2) sensors that can perceive (a) the own body (to be able to become self-aware), (b) own manipulations in the world (to connect to a world in which consciousness resides) as well as (c) manipulations by others (to realize that there are others).

So far, maybe so good, but I am a bit troubled by the subsequent inclusion of explicit symbols and language in the decomposition. If we mean the term "symbol" as being the ability to aggregate and generalize states in the world, then I think I may be for it, but I fear that Butz joins a host of language chauvinists by requiring language in the minimal decomposition of conscious thought:

Language and the tendency to develop a grammar structure might help to shape the representation (leading to further abstractions and yet more general interdependencies in the world such asgeneral cause and effect, somewhat like subject and verb). Finally, it might be the story telling capability, that is, the continuous anticipation working and predicting and guiding future behavior based onthese abstract (sub-)structures in time and space that might lead
to conscious experiences.

Although I've never been able to get a good linguistic answer from the dogs we owned when I was growing up whether they were conscious, their behavior convinced me they were. Moreover, I think consciousness in the Searlean sense (pp. 40-41) of "those states of sentience or awareness that typically begins when we wake up from a dreamless sleep" is available to a whole host of critters beyond those that manipulate or exchange symbols with one another (a growing, but still small class of species).

What is the minimal machinery that gets us to a state of awareness or sentience? Indeed I can't talk about being aware without language, but I don't think I need language to be aware. It is the minimal machinery that we should be concerned with, and I think we need to suspend the need for language to make progress, but include some reflective, reverberating, or awareness elements that interchange external and internal states in some vibrant manner if we are to create systems that are aware. I can't program such an element or elements in Matlab yet, but these subsystems may be more important to awareness than even the capability to aggregate and generalize states.

Friday, April 22, 2005

 

Happy 20th birthday to organized GAs and EC

In 1999, two conferences combined to form the conference currently known as the Genetic and Evolutionary Computation Conference (GECCO). The GP conference was one of those, and the ICGA conference (International Conference for Genetic Algorithms) was the other. ICGA was the first regular conference in the field of evolutionary computation, and the first ICGA was held in 1985 in Pittsburgh, PA on the campus of Carnegie-Mellon University. As a result, I think it is fair to say that the field of genetic algorithms and evolutionary computation will turn 20 this year. Of course, first writings on the subject go back much earlier, but a field requires a regular interchange of ideas, and that process started some 20 years ago this summer. Happy birthday GAs.

 

GA poker players old and new

Alamopoker.com has a post (here) to the abstract of a CEC paper that used evolutionary computation to learn to play poker. This reminded me that the original use of Steve Smith's LS-1 system (the original Pittsburgh genetics-based machine learning system) was to learn poker. That work was completed almost 25 years ago (S.F. Smith, A Learning System Based on Genetic Adaptive Algorithms, Ph.D. Thesis, Computer Science Dept., University of Pittsburgh, Dec 1980)! As always, IlliGAL encourages its readers to gamble responsibly.

 

Blogs will change your business

Hat tip to my colleague Harrison Kim for pointing me to the Business Week article, Blogs Will Change Your Business. The authors also have a new blog on blogging called blogspotting.

Thursday, April 21, 2005

 

DISCUS in the blogosphere

gmtPlus09 picks up on DISCUS here.

 

Is your glass half empty or half full?

I've been working on a major rewrite of my 1995 book, Life Skills and Leadership for Engineers, and I've become interested in Martin Seligman's work on learned helplessness and learned optimism. The research and clininal work is engagingly and brilliantly described in his book Learned Optimism. There are many important thoughts in this research, but a key idea has to do with the relationship between a person's explanatory style and whether they are generally optimistic or pessimistic. Moreover, the work shows that optimism is tied to success in careers, sports, politics, and other walks of life. Although there isn't an explicit study mentioned in the book on academic success, this could be a life-changing read for young academics or assistant professors who might be wondering whether their general outlook on life is helping or holding them back. The book has a number of self-assessment instruments and practical suggestions for changing your explanatory style.

 

Next generation machine consciousness?

One of the symposia that caught my eye at AISB 05 last week was Next Generation approaches to Machine Consciousness: Imagination, Development, Intersubjectivity, and Embodiment. I met a number of the participants of that symposium riding the bus to the train station the first evening, and we had a lively conversation about the prospects for machine consciousness, incIuding a discussion of Dennett, Searle, emergence, the Chinese Room, and so forth, and I wanted to attend their sessions, but was committed to the Conversational Informatics group.

Nonetheless, I've been going through the proceedings, and even a glance at the titles suggests some interesting fare (see here). I've wondered for some time what are the minimal conditions for a computational consciousness and whether the IlliGAL little models analysis and design methodology (see here) might be used to create one. The thought is that consciousness is an emergent property if there ever was one, but an emergent property of a system can be designed when (a) the system is properly decomposed and (b) the elements are of the decomposition are minimally modeled and tuned to yield the desired effect. For example, effective recombinative innovation is an emergent property that was thought to beyond computation (see the Design of Innovation). Why can't we use a similar methodology and create a machine consciousness?

Wednesday, April 20, 2005

 

International Workshop on Learning Classifier Systems

The International Workshop on Learning Classifier Systems is approaching the decision deadline. We are pleased to report we have had 9 full papers and 8 short papers submitted, and that it looks like we are on our way to a successful workshop. More information about the workshop can be found here.

Tuesday, April 19, 2005

 

GAs for macro cell placement

Bob Wall at CS 580 - Computational Science Blog refers his students to Genetic Algorithm for Cell Placement. In this paper, Henrik Esbensen of Aarhus University informs the reader of the popular role of simulated annealing for cell placement in VLSI layouts. When this paper was written GAs had been used elsewhere in cell placement but not for macro cell placement. This is yet another interesting and efficient application of GAs to circuits. Thanks Bob for pointing this out.

As a side note, what a good idea to use a blog for class communication. Easy to setup and good for guided discussions. See David Goldberg's earlier post about corporate vs. academic blogging.

 

Machine learning news ticker

While looking around for more information about TNT Logistics and their use of genetic algorithms last night, I found a nice page for news in machine learning, and genetic and evolutionary computation at cervisia.org. Check it out.

Monday, April 18, 2005

 

TNT Logistics looks to improve distribution center efficiency

Manhattan Associates, Inc., the global leader in providing supply chain execution and optimization solutions, licensed its Labor and Slotting solutions to TNT Logistics North America, Inc., a leading provider of supply chain logistics services. Genetic algorithms are to be used to find the most beneficial and ergonomic placement of pickline items. This is a short quote from the two related press releases (here and here):

TNT Logistics North America has purchased Manhattan Associates' Labor and Slotting solutions to measure the work performed in its distribution centers using detailed information about employees' schedules and activities. The software will enable TNT to analyse the information to facilitate initiatives such as incentive-based pay.

The Slotting Optimisation solution will determine the most beneficial and ergonomic placement of pickline items using genetic algorithms.

Sunday, April 17, 2005

 

Blogging as obsession

Regular readers of this blog may have noticed less frequent blogging on my part over the last few weeks. Some of this is due to two weeklong trips, but I've also cut back considerably when my wife suggested that I was becoming "obsessed" with blogging. In response to that complaint, I'm now blogging every other day (this opens up some free air time for other IB bloggers, hint, hint), but the larger issue is whether others notice blogging impacts on their non-virtual lives and how widespread these effects are among bloggers generally. The feedback and the immediacy of blogging is a rush, but, hey, its just a blog. For those married folk who are unsuccessful in balancing marriage and blogging, here is a post on blogging divorce.

 

Fake paper accepted

Following up on Claudio Lima's nice post on SCIgen, the fake paper generator, apparently the World Multi-Conference on Systems, Cybernetics, and Informatics accepted a randomly generated SCIgen paper on a "non-reviewed" basis. The paper has been returned to the hoax team, but they are now seeking (see here) a volunteer with an accepted paper at the conference in the hopes that the hoax team can present a completely random presentation at the conference. WMCSI was a target of the hoax, because its spam and acceptance policies are both notoriously indiscriminate.

 

Conversational informatics

I had good intentions of live blogging from AISB 2005 in Hatfield, England, but the wireless connection wasn't as a good as I hoped. Nonetheless, the conference was an interesting one, and the symposium I intended was apropos to our work on collaborative systems and innovation support in DISCUS. That symposium, Conversational Informatics for Supporting Social Intelligence & Interaction was organized by Yukiko I. Nakano (RISTEX-JST, Japan) and Toyoaki Nishida (Kyoto University, Japan). Professor Nishida has a book on Google, Dynamic Knowledge Interaction, and much of the work presented at the Symposium originated from his lab. The work presented was diverse, serious, and quite interesting (see the program here), and I hope the symposium continues to thrive and prosper.

Thursday, April 14, 2005

 

Paper Generator

A group of grad students, Jeremy Stribling, Max Krohn, and Dan Aguayo, from the PDOS research group at MIT CSAIL developed a program called SCIgen to randomly generate Computer Science papers:

"SCIgen is a program that generates random Computer Science research papers, including graphs, figures, and citations. It uses a hand-written context-free grammar to form all elements of the papers. Our aim here is to maximize amusement, rather than coherence."

"One useful purpose for such a program is to auto-generate submissions to "fake" conferences; that is, conferences with no quality standards, which exist only to make money. ... Using SCIgen to generate submissions for conferences like this gives us pleasure to no end. In fact, one of our papers was accepted ..."

Yes, it's true! They just got an accepted paper and apparently they are going to the conference to give a random generated talk :) I predict an interesting discussion after the talk... or no discussion at all :-)

Additionally, you can also generate your own paper.

Tuesday, April 12, 2005

 

Rubes

Given the IlliGAL Rube Goldberg sub-theme, it seems worthwhile to blogecho the current boingboing post on the nationwide RG inefficient machine contest.

Monday, April 11, 2005

 

Interceptor uses genetic algorithms for pattern discovery

Interceptor is a search engine marketing company. Today they announced that they developed technology for detecting "click fraud" (click fraud occurs when someone clicks on an online ad with the sole intent to cost the advertiser money). The article about this mentions genetic algorithms as one of the components of their technology:


Inceptor has the only bid management technology that incorporates a powerful, genetic algorithm to detect deep patterns in the data and automatically take action to optimize advertising campaign performance.


There are a number of posts on this blog that relate to marketing, see for example here or here.

Thursday, April 07, 2005

 

Good design resembles nature?

Imaginis fancy guess speculates that
Now that we have enough computer power, we can imitate nature's method as well as its results. Genetic algorithms may let us create things too complex to design in the ordinary sense.
I don't disagree, but I also believe that we need GAs that scale well, competent GAs (see here, here, and here).

The posts discussion of design from nature leads us back to a post on Ingo Rechenberg here.

 

A drumbeat of GAs or a GA of drumbeats?

Geneffects has a post introducing GAs and their use in music. Digging further this itinerant bartender/programmer markets an interesting system called musing that uses interactive genetic algorithms to evolve cool drumbeats. As my 16-year old would say, raw, totally raw.

 

AcademoProfesional Rant

It's not often that I get a chance to have a full scale rant and present academic results in the same place. But a position paper for something like The Grand Challenge in Non-Classical Computation International Workshop affords me that rare opportunity. My position (along with some results from my student Jon Caves) can be found here, in the company of a number of other position papers.

And it's all sponsored by the lovely people at Microsoft Research.

Wednesday, April 06, 2005

 

GECCO-2005 submissions

The submission process is slightly more complex with ACM so please try to allocate enough time to do this final step. Most important points to not forget (strongly biased to latex users) are

1) Latex users - include .bbl file in your .tex source by replacing \bibliography{...} with the body of the .bbl file.

2) Latex users - use sloppy environment starting right after \begin{document} and ending right before \end{document} to reduce the problems with too long lines. This should work almost perfectly. Using sloppy environment locally didn't work with me.

3) Make sure the copyright statement including GECCO-2005 information is in the bottom-left corner of your title page. For latex users, the copyright statement is included in most style files provided by ACM and the following lines must be added in your document:
\conferenceinfo{GECCO'05,} {June 25--29, 2005, Washington, DC, USA.}
\CopyrightYear{2005}
\crdata{1-59593-010-8/05/0006}

4) File names start with f for full papers and with p for posters. See the instructions for examples.

5) Latex users - when generating the .ps file, make sure Type 1 fonts are used and the paper size is letter (use "-P cmz -t letter" in dvips).

6) When using Adobe for creating the pdf file, use the options provided at http://www.sheridanprinting.com/distill_sets/ACM.joboptions

7) Prepare a text-only version of the abstract and references (to cut and paste during the submission). Check international characters and other nonstandard symbols cause this can sometimes cause trouble.

The deadline is April 15, 2005.

Tuesday, April 05, 2005

 

Slurrier's latest list to ponder

Bill Dozier of Notional Slurry (aka the slurrier), has one of his periodic lists of things to ponder here. As per usual, a good bit of it (~63%) is worth pondering.

 

The pickle rides (blogs) again

Franz Dill has some nice posts about computational social science and systems at his blogging home away from IFTF Future Now, The Eponymous Pickle. See here and here.

Monday, April 04, 2005

 

Wharton does GAs

Businesswire reports that Wharton Learning Lab at the University of Pennsylvania's Wharton Business School has released three new simulations. One of them was developed by Professor Steven Kimbrough and is intended for students in Operations and Information Management 101:

The third new simulation is for students in Operations and Information Management 101 and was developed under the faculty leadership of Professor Steven Kimbrough. It is comprised of a two-population genetic algorithm developed as an add-in to Microsoft Excel. Students using this tool learn evolutionary programming techniques for constrained optimization problems, such as those occurring within marketing, manufacturing or logistics.

GAs are a great way to learn about business and competition more generally, and I wonder if those lessons will be conveyed as part of the exercise. Regardless, mainstreaming GAs in the education of future business leaders at a top B-school should be viewed as an important event.

 

GAs and stocks

Sky Yin at Pig Lord's Happy Garden reports on some preliminary results of a portfolio analysis system he designed. He says:

The latest result of optimization test on 10 major Canadian stocks during the past 4 years(based on monthly historical data from Yahoo! Finance) shows that we could get 8%/year higher in return from this system than the "index portfolio"(the weights are determined as how index funds do) including the same stocks.


GAs have been used in determining when to buy or sell which stocks for some years (just do a Google search to see examples - including this paper).

As an example, check out the Trading Solutions financial analysis package that integrates GA technology into their analysis. They report a 60-85% accuracy rate for their trading. Not bad for the stock market.

Sunday, April 03, 2005

 

Behind the scenes of Google

Via ./ I found a lecture by Jeff Dean--distinguished Google’s engineer--at University of Washington. If you are interested in knowing how things are cooked in Google behind the scenes you may want to watch this lecture. This is the abstract of the lecture:


Search is one of the most important applications used on the internet and poses some of the most interesting challenges in computer science. Providing high-quality search requires understanding across a wide range of computer science disciplines. In this program, Jeff Dean of Google describes some of these challenges, discusses applications Google has developed, and highlights systems they've built, including GFS, a large-scale distributed file system, and MapReduce, a library for automatic parallelization and distribution of large-scale computation. He also shares some interesting observations derived from Google's web data.

 

Going to AISB 2005

I'm attending AISB 2005 (Social Intelligence and Interaction in Animals, Robots and Agents) the week after next near London. Specifically, I'm going to give a paper (Mining Social Networks in Message Boards) in the Symposium for Conversational Informatics for Supporting Social Intelligence & Interaction (here). This is a departure from my usual genetic algorithms kind of conference, but the topics look quite interesting and various, and I'm looking forward to attending.

 

Final Four fever in Champaign-Urbana

The Illinois Genetic Algorithms Laboratory (IlliGAL) is located at the University of Illinois at Urbana-Champaign (UIUC), home of the number-one ranked college basketball team in the country. Yesterday after the Illini topped Louisville 72-57 in the opening game of the Final Four series in St. Louis, Missouri, students and other fans poured out into the streets of campustown, creating a crowd estimated between 5000 and 10,000 by police. Despite the large numbers, the crowd was well behaved, and festivities broke up within an hour of the finish of the game. See more here and here.

My last two academic stops were Alabama and Michigan, and I thought I was cured of big-time college athletics, but I must confess that the final-four run has been a lot of fun. Moreover, after years of egocentric big-sports coaching it is refreshing to see a leader like Bruce Weber, who was recently and deservedly named AP Coach of the Year (see article here). The championship game between North Carolina and the Illini is a matchup between two great teams and two great coaches. I know what I'll be doing tomorrow evening.

Friday, April 01, 2005

 

Genetic programming article in BusinessWeek Online

FYI, I found an article As Computers Grow Ever Smarter about genetic programming in BusinessWeek Online. The article was written by Justin Kovac, a finalist in the 2005 Intel Science Talent Search.

This page is powered by Blogger. Isn't yours?