Dream Catcher

This new gadget is very appealing but two things come to my mind:
- why would you want to direct your dream rather than simply remembering them,
- here is a product that sells for 136 buckarus yet has only proven to be effective in 20 percent of the time.
In other news, the use of oxygen mask during sleep for those with sleep apnea is effective in reducing heart problem down the roads according to this finding. The point of that last article is that there is a need for such study so that people overcome the perceived annoyance of having a mask while sleeping: It can save your own life, the study says. What about devising a better way of doing CPAP instead on relying on fear ? (Thanks Cable for the link).

Why the sleeping industry cannot wake up.

In this study of sleep patterns of adolescents in Italy, one can evaluate one of the reason there are not that many customers for a solution to sleeping problems. Out of about 20 percent of the population that self classify itself as poor sleeper, only 2 percent will try to do something about it. In most cases, they will resort to chemical means. wow...

Simon says as music is heard, it begins to make sense

Here is an interesting approach where single musical notes are used as unit word and it is found that musical pieces follow Zipf's law. So as the music progresses, some context and meaning is built into it. Wouldn't that be a new way to compress music ?

Thank god for GPS

This document: High Speed Navigation of Unrehearsed Terrain: Red Team Technology for Grand Challenge 2004 relates the inner workings of the technology of the red team that made it as fas as 7.4 miles into the desert. Interesting.

Hierarchical data object builds itself

Here is a little python script that allows one to define a data structure on the fly. Note that in the class definition there is nothing about name and surname. This is very nice.

GPU programming. Continued.

Here is an interesting approach:PyFX using Python to program a GPU. Very interersting stuff.

Cars as a service

When I was talking to a technical director of a car company, the subject eventually veered of to the concept that in Europe, at some point, people will eventually be buying cars for 5,000 euros. It seems that this already a reality for eastern europe countries but we eventually figured it would be the case for western europe as well after a while. At this price range and below, we imagined the car to become a commodity but we also came to the conclusion that for all intent and purpose, that car could very well come with an engine that the customer could not have access to. The idea is interesting but current cars still require some type of maintenance that makes this idea a little far fetched. And then I read this :
" ...Because the StarRotor engine promises to require little maintenance, it is possible to consider a novel marketing concept in which the engine is leased, rather than sold. The average driver has little interest in the automobile engine. Most really do not want to own an engine; rather, they are only interested in the service it provides. Most drivers are primarily interested in body style and interior furnishings and only become aware of their engine when it requires maintenance. At this point, it becomes a hassle for the driver because he must arrange for alternate transportation during the repairs, which sometimes take more than one day to complete. In contrast, if the driver leased the engine, rather than owning it, he would simply take his vehicle to the auto dealer where the broken engine would be replaced with a working engine, perhaps taking only 30 minutes to do the replacement. Then, the engine repair becomes the responsibility of the automaker..."

This would be a totally different model for a car company. It is pretty obvious that since this is a disruptive technology, it cannot be marketed like your average car and needs to find a specific group of people for which the car is indeed just a service, not a second skin. In Paris, you can rent cars for 9 euros or 10.80 dollars a day, tax included. When Clayton Christensen talked about electric cars in his book on disruptive technology, he made the comment that electric cars being slower than their gasoline counterparts would have to be marketed for a public that was required to drive slowly (teenagers). In this case, the cycle is different and there is no expectation to have a car that runs slower. So, it really looks like this motor should be sold to a different public.

The sleeping industry awakes

It looks to me that the idea of having a sleeping industry is long overdue: Here is room designed that puts customers to sleep. Not only that but "With sleep disorders a rising concern in Japan, the company is betting the elaborate system, which includes a special bed, wide- screen TV and sound-absorbent walls, will cater to a growing market. " The question is, am I too late ?

Location, location, location

Here is a good idea, instead of trying to show different types of statistics of how people are going from one place to the other within an application, what about using concepts we already know such how people move in a city like VisitorVille. In their approach, traffic is represented by people and web sites are represented by buildings. This would definitely be a good way of representing people working in a task manager application such as the one we are developing. Every tower would be a different project, every floor would be a specific area of engineering. More interestingly, the switching time effect between projects could be very nicely represented by the time it takes to go from one building to the other......

Other interesting laws you did not know existed.

I was surfing the web for other laws that i had never heard about before and found the following two: Bradford's law and Lokta's law. Both of them seem related to the social network of scientist publishing and I wonder one can find these same law in implementing this greedy algorithm.

Plouffe's Inverter

These graphics of the number of entries in Plouffe's Inverter are very intriguing. What it plainly says is that nature has an affinity for characteristic numbers that have the first four digits starting with 1000. For instance pi is really 3141 using this approach. The first graph says that there are 6000 constants like Pi that starts with 1000 as opposed to only 800 constants that starts by 9496. The graph represents a total of about 73 millions constants. This is just fascinating, nobody can explain this phenomena, some people think it is because of the base 10 we use to represent numbers. But it goes beyond just constants in the mathematical world. In the social sciences, it is called Benford's Law. Back in 1938, Benford compiled a list of different numbers coming from different sources and found that in every distribution the first digit was always following a distribution where 1 would occur 30 percent of the time. One of the most recent use of this "law" is its use in tracking financial fraud. I have tried it myself on several datasets and it works OK.

New Grand Challenge

Grand Challenge 2005 will take place on October 8th, 2005.

Greedy algorithm could help ranking physical phenomena.

No I do not mean this type of greedy algorithm, where you shut down plants to hike up the prices. Rather, it is a type of algorithm used in networks that is pretty fast and is based on initial sensible solutions. As I was reading the very interesting list of publications by Mark Newmann, I realized I was looking at the other part of the sorting/ranking algorithm that I needed. When one has performed a pairwise estimation of the distance between different objects of a set, what is the best way to plot all these relationships. Sure you could try to plot the correspondance matrix but it would likely confuse everybody but yourself. The article on this fast algorithm devised a way to build a fast ranking of different elements with each other. Let me take an example: let us imagine that you have different photos of a two-phase flow experiment and you want to categorize them in different sets called "bubbly flow", "annular flow". You first compute the "distance" between each pictures. If you have n photos, it means you now have n^2 distances. You then use the algorithm of Newmann highlighted above on the set of distances and you should be able to have different categories of photos representing different catagories of flows. If it works, the interesting thing is that it would be an automated process as opposed to having an human operator qualifying the flow regime.