Thursday, December 01, 2005

Quantifying WOWs and Autism: How many units of WOW can one handle ?


Pierre Baldi and Laurent Itti describe a way to go about quantifying surprise. Their approach is to evaluate the Kullback-Liebler (KL) distance between the Bayesian prior and posterior probability distribution of an event. This seems to work very well.

Using this framework we measure the extent to which humans direct their gaze towards surprising items while watching television and video games. We find that subjects are strongly attracted towards surprising locations, with 72 percent of all human gaze shifts directed towards locations more surprising than the average, a figure which rises to 84 percent when considering only gaze targets simultaneously selected by all subjects

I especially like their "Wow" units.

If you recall, in a previous entry, I mentionned the work of Kevin Pelphrey on gaze following deficiency in autism where he mentions:

On congruent trials, subjects watched as a virtual actor looked towards a checkerboard that appeared in her visual field, confirming the subject's expectation regarding what the actor ‘ought to do’ in this context. On incongruent trials, she looked towards empty space, violating the subject's expectation. Consistent with a prior report from our laboratory that used this task in neurologically normal subjects, ‘errors’ (incongruent trials) evoked more activity in the STS and other brain regions linked to social cognition, indicating a strong effect of intention in typically developing subjects (n = 9). The same brain regions were activated during observation of gaze shifts in subjects with autism (n = 10), but did not differentiate congruent and incongruent trials, indicating that activity in these regions was not modulated by the context of the perceived gaze shift.


So somehow, the autistic subjects are not surprised by the incongruent test. Either the STS is not doing the processing of determining what a surprise is or it is paying too much attention and overwhelmed by details.
Indeed, as shown by Edward Vogel, an assistant professor of cognitive neuroscience at the University of Oregon, it looks like awareness, or "visual working memory," depends on one's ability to filter out irrelevant information.

Until now, it's been assumed that people with high capacity visual working memory had greater storage but actually, it's about the bouncer – a neural mechanism that controls what information gets into awareness," Vogel said.
....
Working with two of his graduate students, Andrew McCollough and Maro Machizawa, Vogel recorded brain activity as people performed computer tasks asking them to remember arrays of colored squares or rectangles. In one experiment, researchers told subjects to hold in mind two red rectangles and ignore two blue ones. Without exception, high-capacity individuals excelled at dismissing blue, but low-capacity individuals held all of the rectangles in mind.

Saturday, November 19, 2005

Server side scientific software as a way to try out an algorithm's popularity ?


I was recently asked about how one could try to test the attractiveness of an algorithm one had devised and see how it was used by users. Generally, people want to be free to toy with an algorithm, so giving them an implementation of an algorithm is fine but how do you do feedback ?

Some of the models I have seen over time include examples such:

* Wavelab where the toolbox allows examples and implementation to be totally readable to your audience
* Beamlab where the toolbox allows examples to be run but hide the content of the algorithm to the user
* The MadMax Optics Matlab toolbox (Multipole method) that can be used for free for a 2-D problem but requires a commercial licence for more complex cases (3D).

All the previous examples use Matlab as a way to remove the complexity of having to deal with different customer's platforms. It certainly does not work well when Matlab versions change. For instance, Carrie Grimes shows us that Matlab Revision 12 and Revision 13 do not handle eigenvectors (eigs.m) the same way. These methods are fine in order to enhance people's understanding of the capabilities of your algorithm, but there are other issues to consider. If you want to improve that same algorithm in order to fit into a specialized area of expertise or several areas of expertises, then the complexities of the previous solution carry a tremendous implementation and platform knowledge burden. In particular, what do you do when you have improved the algorithm ? In most cases, you have to publish a new version of software and hope that your users are tracking your web page or that you were wise enough to have set up a mailing lists and thenhope that all your users still have the same e-mail addresses. Mailing lists handling, in my experience, is a pain and takes away time you should be focusing on the algorithm. Last but not least, you also need to make sure your new algorithm works perfectly on different platforms.

Another way to go about this and avoid the upgrade nightmare is to evaluate the possibility of having a server side software like the Argonne optimization server and provide a GUI that anybody can use. There, computations are performed on somebody else's computer provided that you send the right batch job. An example would be the interactive demo on the simplex method.

Another example with a GUI is Spenvis (a software that provides a way of quantifying radiation doses to spacecrafts while in orbit.) To see how it works, you might need to take a look at it by creating an account. There, you are never running the models on your machine and therefore do not have access to the models except through their reference number. You can run different models and play with many different parameters. If this is an important application, customers may want to pay a licence to use and keep the data on their accounts. They may even ask for a freezed version of the current server side computation. It is also very likely that if a new use of your algorithm arises, an account holder/client might provide a useful input as to what needs to be added to the application to make it appealing to a larger audience. This is not a new idea, Paul Graham lists the reasons why this server-side business is promising in his now famous essay "Beating the averages"

I was initially thinking about implementing something like this server side algorithm for a multilayer algorithm but let's take an example in the space business. If one were to design orbits for spacecrafts performing remote sensing imagery, one could use the different tools offered in the Google Maps API to project these orbits on earth using examples such as these:

* a simple Spacecraft tracking,
* or the shadow of eclipses on the ground,
* or the map of Robotic car race which uses polylines,
* or finally a map of Nuclear detonation which uses an image superimposed on maps.

But why stop at 2-d imagery, when Google Earth provides a 3-d view.

Just generate a klm file and you are done. A way to do your server side algorithm would be to just generate a klm files to be used in Google Earth!

The example of Google Maps or Google Earth serves two purposes:

* With very little capability you can produce output similar to commercial products such as STK thanks to a third party product like Google Maps.
* If you have a good algorithm, it would be of tremendous help to designing a good API.

And don't forget, your initial crowd of users might be students at colleges and universities. You might use the WOW factor to initiate a good following. Wowing examples can be found in the following references ([1], [2],[3], [4])

Other relevant references of interest:
* LAMP: 5
* Ajax: drag and drop site creator, Tutorial on Ajax, here,
and here as well.

Wednesday, November 16, 2005

Comment l'etat americain interagit avec les petites entreprises innovantes (premiere partie)

Ce poste, en francais, vient d'une rencontre a l'un des colloques organise par la SCM ou se trouvait des representants de l'etat Francais, en particulier de la CNER. A la fin du colloque, il semblait assez evident que l'etat francais est incapable d'avoir une relation normale ou une relation (tout court) avec des tres petites entites. Comme il est presque acquis que seules des toutes petites structures sont a l'origine d'innovations impressionantes (Apple, Google,...), il me semblait utile de detailer comment l'etat americain permet a certaines de ces start-ups de tres hautes technologie de fleurir a travers des financements specifiques.

Voici donc une petite revue des structures qui permettent aux petites entreprises américaines d'avoir des contrats avec différentes entités (ministère/agence) du gouvernement américain.

Depuis relativement longtemps, il existe un système de budgets spécifiquement alloués aux petites entreprises grâce à certains programmes. L'un de ces programmes, le plus connu, inclu le SBIR et le STTR qui signifie Small Business Innovation Research Program (SBIR) et Small Business Technology Transfer Program (STTR), il est coordonné par la U.S. Small business administration (ministère des petites entreprises) et a la particularité d'être administré par chacun des ministères/agences qui le finance. Les plus connus sont ceux du département de l'énergie et de la défense ( Département de l'énergie (équivalent du CEA)et le Ministère de la Défense).

Les ministères qui participent à cette activité de R&D à travers des petites entreprises (SBIR/STTR) sont :

..Department of Agriculture
Department of Commerce
Department of Defense
Department of Education
Department of Energy
Department of Health and Human Services
Department of Transportation
Environmental Protection Agency
National Aeronautics and Space Administration
National Science Foundation
..


En particulier, les conditions de définitions pour les contrat SBIRs sont les suivants:

..SBIR Qualifications:

Small businesses must meet certain eligibility criteria to participate in the SBIR program.

American-owned and independently operated
For-profit
Principal researcher employed by business
Company size limited to 500 employees ..


Le programme se résume en trois phases:

..Three-Phase Program:

Following submission of proposals, agencies make SBIR awards based on small business qualification, degree of innovation, technical merit, and future market potential. Small businesses that receive awards or grants then begin a three-phase program.

Phase I is the startup phase. Awards of up to $100,000 for approximately 6 months support exploration of the technical merit or feasibility of an idea or technology.

Phase II awards of up to $750,000, for as many as 2 years, expand Phase I results. During this time, the R&D work is performed and the developer evaluates commercialization potential. Only Phase I award winners are considered for Phase II.

Phase III is the period during which Phase II innovation moves from the laboratory into the marketplace. No SBIR funds support this phase. The small business must find funding in the private sector or other non-SBIR federal agency funding. ..


Les programmes de transfert de technologies (STTR), qui sont utilisés par moins d'administrations, utilisent à peu près les mêmes critères que ceux du SBIR.

Quelques petites remarques liées à ces programmes. La majorité des "fausses" bonnes idées qui ont passé le test de sélection avec les ministères/agences concernés sont en général éliminées à partir de la deuxième phase (car il n'y a en général pas de réel plan de commercialisation ou d'interet d'une autre agence). Il est à noter que la dernière phase ne requiert pas de financement de l'Etat fédéral. Chaque année il y a des dates limites de depots de dossier afin soumettre des propositions de projets. Ces dates semblent être différentes d'une agence à l'autre. Certains contrats SBIR sont aussi soumis à des systèmes de discrimination positive. Dans l'évaluation des propositions, il y a une prise en compte de l'appartenance du chef d'entreprise à des catégories telles que l'appartenance raciale, sexuelle ou le fait d'être un ancien des services militaires. Il est à noter qu'une majorité de ces entreprises est en général créée par des anciens professeurs d'universités ou chercheurs. Il faut savoir que si certaines entreprises reçoivent souvent ce genre de contrat, c'est aussi en particulier, parce qu'elles renouvellent régulièrement leurs équipes, assurant ainsi leur capacité à être au fait de nouvelles recherches prometteuses.

Pour ce qui est de la vérification de la bonne utilisation des fonds au cours de chacune des phases du SBIR ou du STTR, chaque agence fédérale utilise ses propres moyens, très puissants, que sont les administrations des inspecteurs généraux ("Office of the Inspector General"). Chaque agence fédérale dispose d'un inspecteur général et des moyens pour détecter et enquêter sur tout problème lié a l'utilisation des fonds de ces administrations, en particulier ceux du SBIR au sein de ces agences. Au niveau de l'entreprise, le budget de la proposition doit, en général, avoir une catégorie "voyages" de façon à permettre aux chercheurs de présenter leurs résultats à l'agence pendant la duree du contrat.

Un fait intéressant est la comptabilisation publique des statistiques des contrats qui se trouve sur le site de la SBA.

Il y a d'autres mécanismes de financements tel que l'ATP (Advanced Technology Program ) du NIST (National Institute of Standards and Technology) mais celui-ci n'est pas réservé uniquement aux toutes petites entreprises. Ce programme sert de pont entre recherche fondamentale et application industrielle. Enfin, il existe aussi des mécanismes de financement de recherches dans certains états, qui sont trop petits pour intéresser de grandes entreprises tels que les ATP au Texas .

Je n'ai pas mentionné la DARPA car elle fait partie du département de la défense (DoD). La Navale (U.S. Navy) ou l'armée de terre (U.S. Army) ont des programmes SBIR particuliers à leur domaine de compétences. Le fonctionnement de la DARPA est plus spécifique et demanderait un chapitre entier.

Eye Tracking Machines and Autism: A Business Case ? Part deux


This is interesting: Andrew Meltzoff and Rechele Brooks at University of Washington came up with new results on how kids learn how to talk. In this new study they show that...
"... babies who simultaneously followed the eyes of the researcher and made vocalizations when they were 10 or 11 months old understood an average of 337 words at 18 months old while the other babies understood an average of 195 words.

"The sounds they are making are very simple, but some children are looking and making these sounds spontaneously," said Brooks. "They are creating a social interaction or a link. There seems to be something special about the vocalization when they are looking at the toy. They are using social information to pick out what we are focusing on. They can't vocalize words, but they are carefully watching where we are looking. We think they are using social information and getting a boost in figuring out the social and language world together."

"Although the babies are too young to talk to us, those individual babies who are most attuned to our eye gaze are the same babies who pick up language faster more than half a year later," said Meltzoff. "This is a fascinating connection between the social and linguistic world and suggests that language acquisition is supported by preverbal social interaction.

"To do this a baby has an important social regularity to master: follow mom's eyes and you can discover what she is talking about. This study shows that babies first master this social information between 10 and 11 months of age, and it may be no coincidence that there is a language explosion soon thereafter. It is as if babies have broken the code of what mom is talking about and words begin pouring out of the baby to the parents' delight," he said.

The UW researchers are following the same group of babies to see if gaze-following and vocalization at an early age predict increased language understanding and use at 24 and 30 months of age...
"

In the study of autism, scientists generally believe there is a dysfunction of the eye tracking of faces. This is all the more important since, for babies, watching faces and gestures is the elementary process that enables learning as shown in another paper by Rao and Meltzhoff. In short, researchers think that the deficiency in face tracking is responsible for an inability to learn much about social cues and even language yielding to diagnosis of Austim or PSOD.

Yet another interesting finding by Aysenil Belger and Gabriel Dichter reveals that the processing of faces in the brain of autistic kids is similar to that of normals kids. As it turns out, Kevin Pelphrey shows that gaze following deficiency is the issue, not the processing of face information, as I was initially thinking [second part of the SCM talk]. During his investigation, Pelphrey shows where gaze processing is localized in the brain:

using event-related functional MRI (fMRI), we show that in autism, brain regions involved in gaze processing, including the superior temporal sulcus (STS) region, are not sensitive to intentions conveyed by observed gaze shifts.... We conclude that lack of modulation of the STS region by gaze shifts that convey different intentions contributes to the eye gaze processing deficits associated with autism.


It seems there is a increasingly better business case of eye tracking software everyday.

Friday, November 11, 2005

Wednesday, November 09, 2005

Wiping NF1 ?

It is facinating, in a matter of four days, one hears about two drugs that could have a significant impact on neurofibromatosis (NF-1.) First there is this announcement that fumagillin could be used to dramatically shrink tumors associated with the the NF1 condition. And second, that Lovastatin, a drug used for reducing the amount of cholesterol and certain fatty substances in the blood could be used to bring back some fo the major cognitive impairements of people affected by NF1. There are other clinical trials for other drugs. This is a stunning development for what is considered a orphan disease.

Tuesday, November 01, 2005

Visualizing complexity


Through fluctuat.net, I found the following web site that displays many different solutions to displaying the complexity of networks. I am trying to use Google map to map the current bird flu and how it propagates and I am not sure how certain data should be displayed. This should help.

Saturday, October 08, 2005

Wednesday, October 05, 2005

Seeing things the way a robot does

There are interesting videos from DARPA's National Qualification Event before selections are made on Thursday. They can be found on the CarTV site. In this video, Stanford provides insight in the vision system of their entry. Stanford has their own stock of videos.

Wednesday, September 28, 2005

Nuit Blanche 2005, Paris




http://www.pegasus-bridge.com/nuitblanche/

Bonjour si vous parlez Francais, il y a de tres fortes chances que vous connaissiez la Nuit Blanche de Paris. Je me suis essaye a faire cette carte car le site de la ville de Paris est deficient pour ce qui est de representer les evenements qui ne sont pas sponsorises par la dite Mairie. Comme Google Maps ne fait pas les rues/plans pour la France, il y a de fortes chance que ce soit la premiere application du Google maps pour la France. Prevenez moi de toutes erreurs ( je pense qu'il y en a beaucoup sur le chemin vert.) Le site se trouve ici:

http://www.pegasus-bridge.com/nuitblanche/

et met un peu de temps a charger, mais on ne peut plus s'arreter quand on commence a l'utiliser. Vous avez ete avertis. Cliquez sur les icones pour plus d'informations. On peut naviguer sur la carte en cliquant et en déplaçant la souris. Cliquez sur + pour zoomer. Les couleurs font reference au parcours de couleur du site officiel de la Ville de Paris. Les icones/signets blancs sont les evenements non sponsorises par la ville.

The "Nuit Blanche" event in Paris, is a night sponsored by the local government and industry that features all kinds of artists during the night (from 7 pm till 7 am the next morning.) I just toyed with the Google Maps API and did the following maps of the upcoming Nuit Blanche in Paris (this coming Saturday) for two reasons:
  • I wanted to try out the Google Maps API and
  • since France is still not been given any mapping/street capability I thought it would be pretty unique to do this type of thing.

  • I am pretty sure I got it wrong on the green path though. The light and prefered Google map of the Nuit Blanche in Paris is here. The larger version can be found here (it may crash on you). It takes a while to load but is eventually very addictive. you have been warned.

    Tuesday, September 27, 2005

    "...I've seen things you people wouldn't believe..."

    We just had confirmation that Voyager 1 reached
    termination shock last December.

    Cassini took pictures of two Saturn moons two days ago.

    A webcam is installed in the middle of the African desert and you can watch it live.


    For the next three days, you'll be able to go deep underwater in real time.

    Friday, September 23, 2005

    Educating Rita

    Trapping Rita into a corner ?


    What should we make out of the fact that the strike probability of Rita has moved West ?
    On september 21, warm water was all over the Texas coast.
    A day later, that warm water has shifted West.
    Could it be that because of the evacuation of the population from the coast and Houston in particular, warm water rejected into the river stream (houston channel) and on the coast has decreased to the point where other locations on the coast are warmer ? If this is indeed a possible explanation, it is not the best move to have it hit a part of the coast where most refineries are.
    On the other hand, if it means that flooding won't occur in Galveston, I am all for it.

    Thursday, September 22, 2005

    Leaving Houston


    It looks like, there have been only three recorded Category 5 hurricanes before (Labor Day 1935, Camille 1969 and Andrew in 1992) according to NOAA. Traffic reports from Houston at 4:40 AM show most Highways are jammed. In 1900, a Category 4 wiped out Galveston. Sea temperatures for the Gulf show that it is unlikely to decrease in size and strength before it hits the mainland.


    The other benchmark we have for this type of situation in Texas is that of Hurricane Carla in 1961. While in the Gulf it reached Category 5 but hit mainland as a category 3.


    According to the people at the NASA Johnson Space Center (JSC), JSC will be flooded.
    To give an idea of the potential damage, here is the landscape photo of the region before a 22 feet flood (that occured with Carla)...

    ... and here is the same photographs with a 22 feet flood.

    The Brazos Valley was hit with 74 mph winds from this Category 3 Hurricane. Katrina was a category 4 at landfall.

    Wednesday, September 21, 2005

    The perfect non-normal storm ?


    Rita is now a category 5 hurricane. The Gulf of Mexico is very warm. It is supposed to hit south of Houston and then heads its way toward Bryan-College Station. Is it the storm that I mentioned earlier.

    Friday, September 16, 2005

    How many dimensions is a smile worth ?

    Echoing a previous story on how bad face recognition algorithms were, it looks like other methods might do the job other than HLLE.
    One of them is diffusion geometry. A good demonstration of its power can be found in Stephane Lafon's web page. In this demonstration, one discovers that the suite of images on the left can be put in order following one parameter (rotation angle). So the big question is, how many dimensions would enable the classification of a smile ? This is very important because, for instance, it is known that people affected with autism have difficulty to make emotions out....

    Orbital Spokes are really large objects.

    Following up on detecting very large objects, it looks like
    spokes have finally been found on Saturn's rings. They were sought since Voyager took pictures of them 20 years ago.

    I wonder if we could explain these using the Flower constellations of Daniele Mortari ?

    Thursday, September 15, 2005

    Sensitivity to outliers in Autism


    I just found out that somebody had somehow tagged my blog in the social bookmark application del.icio.us under autism. This may be due to the fact that I have written on it in several occasions ( [1] , [2] , [3] )The person who did this is Paul Harrison, so I went to his web page and found a paper he submitted for publication on the theme that Autism is the result of a sensitivity to outliers as discussed in this paper. This is all the more interesting since Wolpert also mentions outliers but in the context of learning sensorimotor functions. An entertaining presentation by Wolpert can be found here (check the Uncertainty on sensorimotor control video.) All this is very uncanny.

    Sunday, September 04, 2005

    Thursday, September 01, 2005

    Katrina People and Location Information Center.


    When hit with a disaster of the importance of Katrina, detailed information is what you want. For every person involved in the catastrophe, there several friends and family who want to know what happened at that location and find out the health of the people impacted. In order to find out about the location, in Katrina's case, people who have left, want to know the state of their belonging. A good idea is to check here and contribute if you have first hand knowledge. For those looking for missing relatives who themselve cannot call you because the cell phones and landlines are dead, here is the place to go ask.

    Non-Normal Perfect Storm approaching


    In a previous entry, I mentioned the issue of non-normality and predicting whether or not a Hurricane's path could be predicted accurately enough to allow for the evacuation of a large metropolitan city like New Orleans. Well it looks we are able to really evaluate non-normal strong convective flow behavior, but these computations seldom can evaluate the cost in human life and the triggering effects they produce.

    Non-normal perfect storms are where seemingly small unrelated elements join together to produce an expected result in proportions unheard of before, are probably the best way to conceive what is about to happen. So while the destruction and damages of Katrina can be witnessed here (before and after picture - move mouse of images) and while the conditions in New Orleans downtown are downright terrible (people are looting to survive) another major effect is looming: oil crisis.


    For instance, at $5.87 a gallon, gas prices are the same as in Europe where one forks about 1.1 euros per liter. This photograph above, while not representative of most U.S. current prices yet, tells you that you now have to fork over 1.2 euros per liter in the U.S., more than some place in Europe. This is just unheard of, since up until three to four years ago, gas prices were about one fourth of the U.S.

    I am also a big believer of small signs. Just last week, I have seen an astounding three references to the SI unit system on the local TV station. It is probably more comforting to hear that the heat wave is at 34 degrees than 100 degrees ! Also mentioning a distance in kilometers or filling a gas tank with liters instead of gallons makes it likely (because they are smaller) these items will become another way of selling products in the future (the liter was in reference to bottles of Dr. Pepper!).

    More recently, the issue of perfect storm was becoming clearer two months ago. With Katrina it becomes crystal clear:


    - There is a similar decrease in oil production in the Gulf due to the Hurricane (Ivan's damages),


    - there is a rise in foreclosures in Houston


    - An oil consultant is thinking, after going through the petroleum engineering literature, that oil estimates in Saudi Arabia cannot be trusted.


    - $70 a barrel.


    - China had an 8.2 % share of global oil consumption last year, and they just have been congratulated by Kazakh officials on the purchase of PetroKazakhstan

    Thursday, August 25, 2005

    Semantically, there are no F5 tornadoes on Mars.


    When Mark Lemmon takes pictures of twisters on Mars, he can't really impress his friends because it is unlikely that a dust devil on Mars will ever get to be called a tornado. See, a tornado is defined by the windspeed of its vortex and more importantly by the destruction it could do to habitation. Since no one lives on Mars and the framerate of the Mars Landers camera is low, it is unlikely that we will know the windspeed of these monster dust devils anytime soon.

    Friday, July 29, 2005

    Meet Planets Lila and 2003 EL61


    Planet Lila was found today and it is bigger than Pluto. But what is more amazing is that there was second announcement yesterday about another planet called 2003 EL61 and they are not the same. What are the odds of finding two planets in the solar system in less than a day ?

    Thursday, July 28, 2005

    A very expensive paint job

    It looks like the paint fragment that



    fell off the main booster tank



    has not affected the belly of Discovery

    Monday, July 25, 2005

    Using a larger pool of talent to refine your models


    When debris flew off Discovery, NASA was surprised about the type of debris that fell off during launch even through they had spent some time to reduce this type of occurence. This is a true statement about how many uncertainties they have to deal with. Another poorly understood process is that of debris falling when they reenter the atmosphere. What is pretty astounding is that the data generally used to detect tornadoes could be used on February 2, 2003 to see real-time the plume made the disintegration of the orbiter.


    While I am pretty sure that these data are being used to refine debris reentry modeling, it is quite exceptional that one can have access to this type of data. With the Python module pyRadar, I am pretty sure an active tornado warning system could be designed somewhat easily using a larger pool of talent.

    Friday, July 15, 2005

    GIS on steroids


    In my first entry in this blog, I made the plea for better and cheaper maps. That was a year and a half ago. Now Google, has come up with two different services that enable pretty much everybody to superimpose additional layer of data on maps and satellite imagery.
    The services are Google Maps and Google Earth. The first service has an API that allows people to use it and add features to the initial service provided by Google. Google Maps Mania lists all these new ideas and the amount of them keep on growing everyday. The space shuttle Discovery lifted off yesterday and one can already see bow different aspect of these services can be used, first one can see where the launchpad was. The same capability can be used to figure out the location of the International Space Station and the Shuttle. Other examples putting meaning and geography together can be found here and there. Fascinating. One can always think of other type of data to be added or how it can be used to tell a story such as that of the Grand Challenge 2004 as John Wiseman did.

    And you thought your day was bad


    In September of last year, I was mentionning the issue of a perfect storm compounding high interest rates, oil platforms, hurricanes and high oil prices. Well, it is not getting any better:

    - First, the Saudis warn of shortfalls as oil hits $61

    - The, we learn that Dennis just tilted a one billion dollar oil platform

    Python is your friend


    Looks like some of the reasons we chose Python for the Grand Challenge race are very similar to the ones Google is using to build its own business.

    WITS, a collection of links to choose wavelets and other related functions.

    Ths WITS lists a pretty complete collection of links to functions such as wavelets, ridgelets, curvelets, contourlets, bandelets and other beamlets. With the increase in the number of families of functions needed to study higher dimensional spaces (above 1), it is a very good idea. Some of these codes are available in matlab.

    Tuesday, June 28, 2005

    Saving Private Rudi

    Captain Miller: He better be worth it. He better go home and cure a disease, or invent a longer-lasting light bulb.


    Rudi Cilibrasi works on the Normalized Compression Distance concept through the use of Google and other tools like Winzip. This is very clever because it uses tools that are already available to everybody while at the same time he uses them to do something that was thought to be impossible. In effect, the data mining and clustering of data in multidimensional space is an issue at the crossroads of many problems of importance nowadays from bioinformatics to robotics, including drug discovery.

    The nice thing about Rudi is that he is making available some of his tools and findings free to whoever wants to use them: his code is on sourceforge (GPL). This is a little bit like what Dave Donoho does at Stanford with his Wavelab software thereby providing proof of scholarship as opposed to the usual article publication which is just a means of advertizing that scholarship.

    Why am I saying these things ? because there is a twist: Rudi is dying. He has Hepitatis C. His parents have already passed away from the disease.

    This is ironic since there is a non trivial chance that one day, a researcher at one of the big pharmaceutical company will use his software, or a derivative of it, in order to discover a drug capable of taking down Hepitatis C. It is also very likely that the product of this discovery will belong to the drug company's Intellectual Property (IP) and that Rudi may be restricted to have access to this drug because he is on the wrong continent or country as has been the case in his story.

    In our war on diseases, it would be a damm shame to lose Rudi since he may even be part of the solution directly or indirectly.

    Please, sign the petition, let's Save Private Rudi.

    Friday, June 17, 2005

    Bob, I would not call it "co-existing"

    Via BoingBoing, I found this tidbit about how Australia's megafauna coexisted with humans:

    New research conducted by Australian and British scientists reveals that in fact humans and megafauna, such as gigantic three tonne wombat-like creatures, a ferocious marsupial "lion" and the world's all-time biggest lizard, may have co-existed for around 15 000 years.


    Don't get me wrong, I am not objecting to the fact that they were not living at the same time, rather the word "co" in latin translates into "with". I'd rather qualify the situation as humans surviving against some pretty tough odds.

    Wednesday, June 08, 2005

    Catch my gamma rays if you can.

    NIAC just released their new grants. One of them is focused on how to harvest antimatter. Why antimatter ? well because everybody knows that if you use some antimatter with its matter equivalent, they annihilate with each other releasing a large amount of energy. Why do we care ? well, the story goes that with a lot of energy, you can go anywhere in space. In essence, once you get antimatter, you're ready for some Star Treck action.

    The main problem is that the last statement is just plain wrong. In order to go anywhere in space, you need to eject mass from your spaceship. You know, inertia, if something goes one way, the rest goes the other way. Hence, in order to have a rapid spaceship, you have to efficiently eject mass as rapidily as possible in the opposite direction. The only way we know how to rapidily eject matter is to heat it up. The warmer the heat, the higher the ejection speed of the fuel and therefore, the higher the speed of your rocket. This is why chemical rockets are less efficient than nuclear rockets.

    In the case of the "antimatter" devices proposed, the large amount of energy released is just massless photons, high energy photons, but massless photons nonetheless. How do you suppose large masses of matter can be heated up and ejected when the fuel is made up of massless photons ? Well if they remain photons, it won't help therefore you need a conversion process enabling matter to pick up the energy of the massless photons instead. And here lies the problem, the cross section of gases to high energy photons is very very small. In effect, physically, it is so poor that it does not compete seriously with any existing and tested nuclear rocket concepts.

    Wednesday, June 01, 2005

    Eye Tracking Machines and Autism: A Business Case ?

    While searching for a gaze tracking hardware/software in order to better understand drivers/driving for our entry in the Grand Challenge Race, I was talking to one of the manufacturers of eye tracking software (Eyelink, SMI and Seeing Machine) and found out that there was not much awareness with regards to issues of autism and some of the solution devised for the eye tracking business.

    In order to make it clear that the two are related, I used the paper of Kevin Pelphrey, Sasson, Reznick and Paul published in the Journal of Autism and Developmental Disorders in 2002 and extracted one of the most compelling picture that one needs to see in order to understand some of the gaze tracking issues of autistic people.


    Eye Tracking Deficiency in Autism Posted by Hello

    Obviously, the brain is not tracking the most interesting part of the face and therefore, very little can be gathered with respects to other people's emotion. Learning may as well be disrupted since face tracking is important in learning from imitation.

    Some recent studies seem to show that one can detect autism or related cognitive deficiencies by watching videos made by families at age 3 to 6 months. This is very important since, way before going into some biochemical or more intrusive means of detection, I am sure that a video based system, like the one sold by these companies, has some potentiality for cognitive deficiency detection. One needs to realize that there is really no detection mechanism for autism before the age of two and it is generally very late at that stage.

    There is a business case to be made especially if you consider that some families are willing to pluck $75,000 of their own money over one year to do something about the autism condition of their kids. More specifically, while it would be extremely difficult to use this technology as a standard test, I am sure however, that families that already have had autistic kids would want to know if their newborn siblings have the same problems. And in the event that they indeed have a similar problem, they would look into working on the imitiation learning business much earlier than at the age of two.

    Monday, May 30, 2005

    Burning Both Ends of a Space Candle

  • According to Nasawatch, 27 percent candles won't lit even though Russian specialists say it is more like 20 percent.
  • There are 82 candles remaining which gives us about 22 candles that won't lit.
  • 60 candles are left to insure enough oxygen inside the International Space Station.
  • 2 candles are used every day.
  • There are only 30 days left.
  • Today is May 30th.
  • 30 days from now is June 29th (Even using the figure of 20 percent will buy four days)
  • The next Shuttle scheduled to go up will do so no earlier than July 13th.
    So the question is: will the International Space Station be mothballed for a minimum of fifteen days until the next Shuttle comes in ?
  • Friday, May 27, 2005

    Did Google Image catch a lenticular cloud ?

    A lenticular cloud looks like this. It certainly looks more like this than some restricted area blanking.

    What is the plague's ping ?

    If bacteriums use propellers to move around, wouldn't it be likely that we can detect them (and even differentiate them from each other) by listening to noise induced by the propellers like we do for submarines ?

    Wednesday, May 25, 2005

    Classifying images

    In the Grand Challenge race, one of our approach to make classification of images for the purpose of road following as well as obstacle avoidance is based on defining some type of distance between drivable and non-drivable scenes. There are many different projection techniques available in the literature. Let us take the example of the "zipping-classifier" mechanism and how we are using it for obstacle avoidance. This technique uses a compressor like Winzip or 7-zip to produce some type of distance between files (here images.) A very nice description of it can be read here. The whole theory is based on Minimum Description Length. An example of a naive classification made on pictures we took at the site visit in San Antonio are shown below. In the interest of reproducibility, we used CompLearn to produce this tree. (Rudi Cilibrasi is the developer CompLearn.) The tree clearly allows the segmentation between open areas and the ones with trash cans. We did no processing to the pictures. Obviously, our approach is much more specific in defining this distance between a "good" road and a not-so-good one.

    classification of trash can obstacle Posted by Hello
    Thank you Cable Kurwitz for providing us with the tree.

    More on the rest of our approach can be found here.

    Friday, May 06, 2005

    Finding your chute when you have already landed, well crashed

    Here is a very nice account on how much work it takes to find a very small feature on Mars.

    GPU and Incomplete Signals that save lives

    Well, it looks like it is now possible to reconstruct PET scans with a graphics card as shown by Fang Xu and Klaus Mueller. Now if we could merge this approach with that of Candes, Tao and Romberg (figure 6), I think we are going somewhere.

    Wednesday, May 04, 2005

    Self serving thoughts or the way of the Lizard ?

    When Dan Moniz reports on Paul Graham's talk, he mentions the following tidbit
    Graham had gone around to several large tech companies (Google, Yahoo!, Microsoft, etc.) and asked them 'if you could hire someone who had tried and failed to launch a start-up company, or someone who had graduated college and went to go work as a developer at a big company (e.g. Microsoft), which would you prefer?' According to him, every single company responded that they would be more interested in the candidate who had tried and failed at a start-up; he included a verbatim quote from the person he talked to at Yahoo! who said he could quote him.

    My experience has mostly been the reverse, but this survey reminds me of how NOT to do a survey. Clotaire Rapaille shows you how to find the inner lizard in all of us (see The science of selling) and how answering such a survey will not give you the right answer. Corporations do not take risks, they prefer youngsters for a reason, their inner lizards tell them so. What do I mean by inner lizard ? Here is an explantion from the persuader's script:

    DOUGLAS RUSHKOFF: [voice-over] Of course, it's impossible to know if Rapaille's excursions through the collective unconscious really uncover what drives us, whether to Boeing airplanes or any other product. But even if he is onto something, you have to wonder about the net effect of reducing us to our most primal impulses.

    [on camera] What about the environment? If the lizard wants the Hummer–

    CLOTAIRE RAPAILLE: Right.

    DOUGLAS RUSHKOFF: –then– and the lizard's not going to listen to the environmentalist–

    CLOTAIRE RAPAILLE: Right.

    DOUGLAS RUSHKOFF: –then isn't it our job, as aware people, to get the reptile to shut up and appeal to the cortex, to appeal to the mammal?

    CLOTAIRE RAPAILLE: Now, you see, the problem is here, is that, if you think, right, the people who want to do good not always do good, all right? So the people that want to do good – for example, let's say, OK, we need to make smaller cars, right, to protect the environment. Then nobody buys the smaller car. Why? Because they're too small. So then the result is they go into trucks.

    DOUGLAS RUSHKOFF: [voice-over] Looks like I'm not going to win this one. After all, it's hard to argue against the reptilian brain.

    CLOTAIRE RAPAILLE: We have to understand the unspoken needs of the people. It works. Good marketing research works. When we say it works, it mean that marketers understand the real need of the customers – sometime unspoken – and they deliver. "Give me what I want."


    For a corporation, which has processes in place and acting in the status quo and trying to reduce its variation around its normal operation, the "Give me what I want" translates into "Give me youngsters I can impress and mold."

    Tuesday, May 03, 2005

    The Meaningful Information Architect

    I was pleasantly surprised to see how the description of Karl Rove put an emphasis on information mining. It is one thing to gather data, it is a different one to make sense of it. What I want to know is: why does watching CSI make you a republican ?

    Sunday, May 01, 2005

    Preconceived notions

    Maybe Eileen Collins is a Roller-Coaster phobic but so are many astronauts. It is well known that many in the astronaut corps do not like to fly inside the KC-135. One of the reason is that space sickness experienced while in orbit has nothing to do with the type of problems one can develop in the KC-135. In effect, it is known that people who are not sick in the plane can become very sick in the space shuttle.

    Friday, April 29, 2005

    Earth sucks it up

    This article is troubling. Not because it is new, but because most modeling of Earth weather circulation assume thermal equilibrium, which as it stands, does not seem to be a good assumption.

    Wednesday, April 13, 2005

    Look out for big lone stars

    As I was hinting in a previous entry, the flower constellations of Daniele Mortari could also be used to construct large "objects" in space in order for us to give signals to ETs. If you take a look at some of the constellations that can be built, it does not take much time to realize that these are made by intelligent people, with a sense of art at that. In this article, Luc Arnold proposes to find ET life by looking at large objects, let us hope there is another Daniele Mortari, several parsecs away who has made the same discovery.

    There is a good reason as to why these figures would be better detected with interferometry systems. One of the interesting feature of these constellations is that the human brain looks at them and figures they have a shape because they see three of more satellites as if they are always be forming a line. This is mainly an averaging effect but it is striking (Daniele can produce a constellation that reproduces the contour of a star, imagine that, a star looking at a star!) Since interferometric systems do make an average view (they have to collect enough light), we are also likely to see these lines in the interferometric data. And the best way to find them, will be use the curvelet transform. Since interferometric systems are also sparse, in term of light collection, it is likely that we will have incomplete data, Emmanuel Candes who has been working on curvelets, is also now working on incomplete fourier ensembles and he is aware of interferometric systems as witnessed in his recent paper entitled "Stable Signal Recovery from Incomplete and Inaccurate Measurements"

    ...Fourier ensemble. Suppose now that A is obtained by selecting p rows from the n×n discrete Fourier transform and renormalizing the columns so that they are unitnormed. If the rows are selected at random, the condition for Theorem 1 holds with overwhelming probability for S  C · p/(log n)3 [4]. This case is of special interest as reconstructing a digital signal or image from incomplete Fourier data is an important inverse problem with applications in biomedical imaging (MRI and tomography), Astrophysics (interferometric imaging), and geophysical exploration.


    It looks like, it is just a matter of time before we go out on a search for a big lone star, you know the five legged one...

    Tuesday, April 12, 2005

    Modeling extinction

    This article on exploding star and its relation to extinction led me to read the fascinating Model of Extinction: A Review by Newman and Palmer. I always wondered about periodicity of astronomical event and their influence on life on earth. It looks like that while the astronomical hypothesis is interesting, it does not seem to be the prevalent means of extinction.

    Friday, April 08, 2005

    Basis Pursuit

    Here is a very interesting paper on the use of Basis Pursuit to find a correlation between different parametric models and data from time dependent PET scans. The nice thing that makes it clear is that there is no need for an ad-hoc procedure to make sure that some coefficients are non-negative (non physical.) We have this kind of problem in neutron transport. In that area, ways to deal with this type of problems are called flux fixing.

    Monday, April 04, 2005

    Divergence-free colliding thoughts

    Have you ever read a subject of interest that was barely relevant to what you were doing right then and found out it was the solution of somebody else's problem. This just happened to me this week-end. I received the NA Digest newletters with a question on divergence free and interpolating functions. Well it just so happens that I found recently an excellent resource on campus here at Texas A&M University. They have begun digitizing all the M.S and Ph.D theses produced by this fine institution, they are available at: https://txspace.tamu.edu/.In this database, I found the doctoral thesis of Svenja Lowitzsch on Approximation and interpolation employing divergence-free radial basis functions with applications.

    Friday, April 01, 2005

    It's full of stars...and comets

    If you look closer to this picture taken yesterday from Huygens, you'll see a bunch of comets. Now the question I have is, does anybody tries to find new comets from these pictures ?

    Thursday, March 31, 2005

    Test Site Visit


    This is the site at the Southwest Research Institute where we will be demonstrating our vehicle.  Posted by Hello

    Printfriendly