Hip Hip ... I bought a new cell phone - and its much more fun than I anticipated. In fact starting with a horrible lousy Samsung a few years ago, I upgraded considerably when I bought a Nokia E55. But it didnt had a camera and I never really got to explore it for real. So this time I went for a decent camera and have bought a Nokia N73 music edition. And it seems to be a lucky choice.

Ok, If you are as impatient as I, you are now thinking "get the bull over with and give me some facts". Ok here they are in two lists:

Bad things

* Reactions are a little slow
* The phone is somewhat large
* Keys with letters/digits small
* Ringtones too fancy

Good things

* Joystick and navigation-keys are fine
* Nice camera
* Good sound
* Large screen
* A lot of fun gadgetlike effects which simply works

There are a few comments, though, for the curious:

Move data. So i had a lot of things at the old E55, but setting the two up with bluetooth was easy. then, using "Move data" it moved almos tall data from the old to the new phone. Exceptions was "templates" in messages, my goosync settings and my Gmail program. Well.

Buttons. The number-keys 0-1-...-9 at the center are rather small. I have to aim carefully or use the finger nails when writing and I am not even having particularly large hands. I cant imagine a person with Parkinsons or a construction worker dialing anything with ease. The central joystick, on the other hand, and the four important buttons above the number-keys are fairly large and easy to use. This is important because much of the activity with the phone is using these five buttons. So mixed results on the button size, I guess.

Speed. For some reason this it has almost always low priority to make phones respond quickly to whatever you want them to do. The N73 with Symbion60 feels similar to E55, perhabs even at bit better, but still you will have to wait for up to a second on some things. It drives me mad, but apparently there are not enough people feeling the same way for Nokia to change it. I am not sure what the goal should be, if 0.1 sec is fast enough - I just want the feeling of instant reaction. It is the same with the camera; as with many other ordinary digital cameras, it takes a while from I push the button until the picture is actually taken - and meanwhile,the Loch Ness monster has dived again, your mothe rstopped smiling or the party is over and people has left. Sigh.

Camera. So the speed in the camera is not cool, but many other things are. It is 3.2 MPixel (a spec which sounds stupid in about a week) and have many adjustments for shuttertime etc. And a real autofocus. Many pictures are good, but there are still some of these over-contrasted disfocused ones ...

Music. It worked easily hooking the phone up with the PC using USB cable and copy the music to the phone using the Nokia PC Suite. i could in a hurry move an album with Trentemoeller and rush off to the concert without getting stressed out over technology almost working. Reasonably easy and nice. Of course the music is stored on the mini SD card so there is room for max 2 GB. Decent, I guess.

Radio. The build in radio, which requires use of the headset/plugins worked perfectly and complete effortless. Good experience indeed!

Ringtones. This is a standard complain I have: Why the hell are there no ordinary phone-like sounding ringtones. Does everybody HAVE to have a stupid song or sound clip? I visited some websites to find something I could download, but no. Out of 30,000+ ringtones (they claimed), there were not a soingle one sound like, e.g. a electrical phone. The ringtone "Low" which was on the E55 was no longer there, so now I am stuck on that issue.

The first machine learning related IAPR workshop on cognitive information processing CIP2008 is announced to be held in Greece this summer. From the website:

"The first workshop on Cognitive Information Processing aims at bringing together researchers from the machine learning, pattern recognition, signal processing and communications communities in an effort to promote and encourage cross-fertilization of ideas and tools. The focus of CIP 2008 workshop is on Cognitive Radio."

Plenary speakers:

• Prof. Simon Haykin (McMaster University, Canada)


• Prof. Timo Honkela (Helsinki University of Technology, Finland)


• Prof. Jose Principe (University of Florida,U.S.A.)


• Prof. Ali Sayed (University of California LA,U.S.A)


• Prof. Bernhard Scholkopf (Max Planck Institute,Germany)


• Prof. Naftali Tishby (The Hebrew University,Israel)

Important dates:

• Submission of full paper: January 5, 2008
• Notification of acceptance: March 5, 2008
• Camera-ready paper: March 31, 2008

• Learning theory and modeling
• Bayesian learning and models
• Information theoretic learning
• Graphical and kernel methods
• Adaptive learning algorithms
• Ensembles: committees, mixtures, boosting, etc.
• Data representation and analysis, PCA, ICA, CCA, etc.
• Other related topics

• Cognitive radio
• Cognitive component analysis
• Cognitive dynamical systems
• Distributed, cooperative and adaptive processing
• Other related topics

In a remarkable article in Journal of Machine Learning Research (JMLR), a group of authors advertise an open source model for peer-reviewed software publication in the scientific machine learning community. The authors argue that the lack of scientific recognition for good software motivates the development of algorithms rather than implementations and that this hinders the true potential of the learning algorithms.

The article is remarkable because of the content of course, but also because JMLR is normally a strictly technical journal and the authors are high-profiled within the community. The complete author list is:

• Soren Sonnenburg, FIRST


• Mikio L. Braun, TU Berlin


• Cheng Soon Ong, Friedrich Miescher Lab


• Samy Bengio, Google


• Leon Bottou, NEC Lab


• Geoffrey Holmes, Univ Waikato


• Yann LeCun, NY Univ


• Klaus-Robert Muller, TU Berlin


• Fernando Pereira, Univ Pennsylvania


• Carl Edward Rasmussen, Cambridge


• Gunnar Ratsch, Friedrich Miescher Lab


• Bernhard Scholkopf, Max Planck


• Alexander Smola, Australia National Univ


• Pascal Vincent, Univ Montreal


• Jason Weston, NEC Lab


• Robert C. Williamson, Australia National Univ

The tenth machine learnig summer school has been announced to be held in Kiola, Australia at the dates March 3 to March 14. The MLSS is a summerschool focused on machine learning - a description from the website:

"The Summer School is intended for students and researchers alike, who are interested in Machine Learning. Its goal is to present some of the topics which are at the core of modern Machine Learning, from fundamentals to state-of-the-art practice. The speakers are leading experts in their field who talk with enthusiasm about their subjects."

Announced speakers are

• Avrim Blum, "Online Learning, Regret Minimization, and Game Theory"
• Wray Buntine, "Latent Variable Models for Document Analysis"
• Tiberio Caetano, "Inference in Graphical Models"
• Nando de Freitas, "Monte Carlo Simulation for Statistical Inference, Model Selection and Decision Making"
• Marcus Hutter, "Introduction/Overview of Statistical Machine Learning"
• Rao Kotagiri, "Data mining"
• Dale Schuurmans, "(Un)Supervised Learning"
• Alex Smola, "Kernel methods and Support Vector Machines"
• Vishy Vishwanathan, "Machine Learning Laboratory"

USA Today is reporting a story on how RFID tags will solve the infinite mess in our homes:

"Voilà! Major parenthood problem solved. Then you could spend the 6.5 hours per week normally devoted to picking up after your kids to drinking martinis on the back deck. What could possibly be a better benefit of RFID?"

Well the story has an air of ironic distance, but nevertheless loyal to the ideas of Kevin Ashton, former leader of the Auto-ID Center, setting the standard for RFID tags. Thanks to Sita Vickery for this one.

In the movie "The Number 23", the main character Walter Sparrow is going crazy thinking about the number 23. Seemingly, the number is everywhere and has a special link to his life. If Mr Sparrow was trained in machine learning or statistics, he would have recognized that he was over-interpreting due to either overfitting or data set selection. Here is how:

Overfitting
Overfitting is to adjust the model to the individual data points of a data set instead of the hidden dynamic. As an example in the line of historical conspiracy, one can select a data set of three well-known tyrans acting in the 20th century:

• Pol Pot
• Hitler
• Stalin

Is there a mystical link? Oh yes, indeed:

1) Simply count the number of letters in each name: The number of letters is 6 for each of them, together forming 666; the biblical number of the beast.

2) Less obvious is number relations: Take the sum of digits from their birth dates and subtract the smallest non-prime number not present, if the digit ”0” is not in the birth date. The result of all of them is 32 – in the movie a sort of ”inverse” of 23.

The above is of course non-sense. The birth dates have su m of digits 32, 32, and 36, and making them end up on same number takes some effort. In fact, I had to carefully sinpect all three dates to look for something I could subtract from 36 (Stalin) but not from 32 (Pol Pot and Hitler). And it is this careful tweeking which is the mistake: If one can choose the system freely given three dates, any three dates can give any number in the end – thus no special coincident or conspiracy.

Data set selection
The above happens all the time, but he case of data set selection is in fact more common. Data set selection is having a (more or less) fixed model, but choosing the data points of the data set to fit the model. In everyday life this is what happens when people get an idea about something and then notice whenever reality fits with the idea. The problem is that they do not pay the same amount of attention to situations where the idea does not apply.

In the case of death and tyrans, one can instead fix the model such that date plus month equals 23 and then search the world history for exciting events. Then one finds e.g.

• December 11, 1994 Russian President Boris Yeltsin orders Russian forces into Chechnya
• November 12, 1927 Trotsky expelled from Soviet CP; Stalin becomes undisputed dictator
• October 13, 1864 Battle at Darbytown Road Virginia (337 casualties)
• September 14, 1812 Napoleon enters Moscow
• August 15, 1961 East German authorities begin building the Berlin Wall
• July 16, 1918 In Yekaterinburg, Russia, Czar Nicholas II and his family are executed by the Bolsheviks
• June 17, 1992 Slaughtering by Inkhata-followers at Boipatong, South Africa, kills 42
• May 18, 1940 German troops conquer Brussels
• April 19, 1941 Bulgarian troops invade Macedonia
• March 20, 1995 Nerve gas attach on Tokyo subway
• February 21, 1916 Battle of Verdun in WW I begins (1 million casualties)
• January 22, 1941 1st mass killing of Jews in Romania

Summary
In summary, it is equally deceptive to adjust the model/pattern too freely or to select data points which maches the pattern. Both roads lead to wrong conclusions. One needs to balance the data set and the model in just the right way and statistics and machine learning have two different but equally good ways to do that.