Entries from Stephan Paukner

Finally! Canon announced the successor of the EOS 30D today: The Canon EOS 40D. Unfortunately, it doesn’t have a full frame sensor, what I had wished secretly, but it seems that these haven’t reached a reasonable price range yet. So, it still crops the view with a factor of 1.6, but I’ll try to stick to the better EF lenses anyway rather than just using EF-S. Nevertheless, the main enhancements to the 30D are:

• 10 MP instead of 8 MP,
• 14-bit processing instead of 12-bit,
• DIGIC III processor instead of DIGIC II,
• Dust reduction by filter vibration instead of nothing,
• Interchangeable viewfinder screen instead of a fixed one,
• ISO sensitivity icon in the viewfinder instead of no indication,
• 3.0” TFT LCD instead of 2.5”,
• LCD Live View,
• Multiple exposures on mirror lock-up,
• 3 custom user modes instead of none,
• Continuous shooting with a rate of 6.5 fps instead of 5.0 fps, and a burst of 75 JPEG or 17 RAW images instead of 30 JPEG or 11 RAW images.

Wanna have! This one will be it. Expect it in my possession at the winter family days (vulgo Christmas) this year.

Update 08/21: Some Austrian vendors already list the EOS 40D at a price of €1,299. My vendor of choice will be Foto-Video Sobotka in Vienna, as I already bought my PowerShot A710 IS there and I got some vouchers for their shop. My first lens might be the Canon EF 24-105mm f/4L IS USM, what Sobotka lists at a price of €999.

For the first time it really seems that I can keep my time plan. I try to reach page 32 today, and if I manage to keep writing one page per day, I will reach the half of the average thesis page number this week, and will indeed manage to be finished with the beginning of October. The theses at my institute indeed range over a number of about 70 pages.

I met with HGFei on Friday, and he said he can fully support my will to be finished soon. There won’t be a serious research or new results in my thesis (except for how the Gabor transform can be computed more easily for separable 2D atoms), but only a summary and visualization of the known results in the 2D case, where of course will be no surprises. Serious research is reserved for dissertations anyway. He wants the results to be suitable for talks or demos.

I proceeded with Chapter 4 about digital images and 2D frequencies rather than with Chapter 3 about finite discrete GA, as it isn’t that mathematical and I wanted to read more on the corresponding things before. I’ll soon have it finished, and maybe I’ll manage to finish the finite discrete GA in the first September week, what is 2 weeks from today.

I already got a little time panic, as I noticed that I wouldn’t make it within September. I’ll need the whole month of September, too. A problem is that Doris wanted to go on a two-week holiday. As she has to take a course anyway, we delayed it to the beginning of October. After that I only want to have to do cosmetic stuff, and no content anymore. I want to have it printed and handed in by the mid of October. And already at the mid or the end of November, there should be the Master exam. I want to be finished by my 30th birthday. This. Must. Work. I can’t afford needing more time, as January 2008 is the last possibility, otherwise I’d have to pay €6,000 back.

On the content: I looked at multi-window Gabor frames, and M. Dörfler’s PhD thesis is a good source. Maybe I can apply her results 1:1 to images, as natural images also dominate in the lower frequencies. The lower frequencies are “area-producing”, already determining the image to a very high degree. Therefore small windows are needed densely in the lower frequency area, resulting in a dense covering here, incorporating many different orientations, yielding only a rough coverage of the image space. The higher frequencies are “contrast-producing” or “border-producing”. Frequency coverage doesn’t have to be that dense here, the exact frequencies are not that interesting, but their location is important. So, a multi-window Gabor system could be of similar type in 2D as it had been taken for 1D music signals by Dörfler.

But multi-window systems shall appear rather late in my thesis. First there are a lot more other things to mention. I made a rough table of contents, giving me a good lead:

• Gabor Frames in as Section 2.3
• Chapter 3: Finite-dimensional discrete Gabor Analysis
• TF-matrices, stuff, frames in
• Gabor matrix
• GA on finite groups and general lattices
• Dual atom on general sampling sets
• Chapter 4: Frequency behavior of digital images (nix Gabor, nix shifts)
• Digital representation of images, RGB, matrix, values
• Understanding 2D frequencies, tensor product of 1D freq
• FFT2, low frequencies dominate, point to multi-windows
• Chapter 5: Image representation by Gabor expansion (Gabor stuff, experiments)
• Atoms can be separable or non-separable
• 2D PF-shifts of an atom, 4D position-freq space
• Separable atom: 2D-dual is tensor of 1D-duals
• Huge frame matrices, applying TF-matrices to image matrices for sep-atoms, proof
• Separable atom, separable lattice, separable dual
• Separable atom, non-separable lattices (1 dim only, both dims), dual
• Non-separable atom, isomorphism 1D-2D, various lattices, dual
• Unassigned
• Multi-window Gabor frames, similar to music signals
• Biological vision

Chapters 3 and 4 should be possible within August, and the rest will be in September.

It is currently completely open if there really will be some “real world” applications of GA to image processing, like serious deblurring, denoising or compression. I’d have to compare it to existing methods, actually. Currently I only plan to compare the various duals and do some thresholding of the Gabor coefficients. If there really will be some multi-window stuff, then maybe only for separable atoms, as I don’t know how I could check the frame quality otherwise.

Subject: xu Hallo

Hallo {, MAILTO_USERNAME}
[...]
Sie wollen, der Ihres “ Gurke “ GROSS & STARK geworden ist, als ein STEIN?
[...]
Tippen Sie Browser (ohne Plätze) Presse Einträgt ein.
[...]

I’ve written 22 pages covering the basics of time-frequency analysis, walking a path from FT to STFT over frames in Hilbert spaces to Gabor frames in L². I actually wanted to have that done within June. Currently I’m tackling with how to include graphics the right way. Next I want to look at multi-window Gabor analysis and think about how I should proceed to finite-dimensional discrete GA and non-separable lattices. Maybe I can start applying this to images in one month. I definitely want to have my thesis finished within September.

I wanted to do a regular upgrade of xserver-xorg-core in Debian ‘testing’. Unfortunately, dpkg complained about a file ./usr/lib/xorg/modules/extensions/libglx.so not being present, although Nvidia’s installer had placed it right there and it was still present. Because of that broken upgrade my X server couldn’t start anymore. Removal of Nvidia or xorg didn’t change anything on this situation. With Google I saw that someone found a solution by installing nvidia-glx-legacy temporarily. Here’s what to do:

1. Remove Nvidia by calling sh /path/to/NVIDIA-$SOMETHING --uninstall.
2. Install nvidia-glx-legacy. This package depends on a linux-image-2.6.18-n-486, what will therefore be installed and will replace your current symlinks /vmlinuz and /initrd.img.
3. Install/upgrade xserver-xorg-core. (The top-dependency would be xorg).
4. Remove nvidia-glx-legacy and the above mentioned linux-image-$SOMETHING again and purge their configuration files (Key “_” in aptitude).
5. Restore your original symlinks /vmlinuz[.old] and /initrd.img[.old]. You’ll have to run lilo. Run lilo. Did I mention to run lilo?
6. Reinstall Nvidia and restart gdm.

Recently, I had another issue with the Nvidia installer: It didn’t work anymore due to the activated paravirtualization feature in Debian’s default kernel what conflicts with the GPL-incompatible module nvidia.ko. Here’s a solution I found in the web:

1. Install linux-source-2.6.xx-n-686.
2. Uncompress /usr/src/linux-source-2.6.xx-n-686.tar.bz2.
3. Delete the symlink /lib/modules/2.6.xx-n-686/build and make it new by ln -s /usr/src/linux-source-2.6.xx /lib/modules/2.6.xx-n-686/build.
4. Copy .config from headers to sources by cp /usr/src/linux-headers-2.6.xx-n-686/.config /usr/src/linux-source-2.6.xx.
5. In the sources dir make menuconfig and disable paravirtualization in section ‘processor features’.
6. make prepare.
7. make scripts.
8. Now compile and install Nvidia with its own installer.

For the sake of documentation, I list the enhancements I configured for SpamAssassin since February:

• I use SA’s internal sa-update script nightly to update the standard rules that are changed between the releases of new versions. The standard channel is updates.spamassassin.org. The files automatically go into /var/lib/spamassassin/$VERSION and Debian automatically finds it, as can be seen from a call of spamassassin -D --lint.
• In addition to the previous, I call sa-compile to compile the body rulesets into binary form. This makes checking of body rules more efficient. To enable SA using these binary rules in /var/lib/spamassassin/compiled/$VERSION, I have to activate the Rule2XSBody plugin in /etc/spamassassin/v320.pre.
• I don’t update the SARE rules against stock spam manually anymore, but also use the provided sa-update channel for it.

Recently, PDF spam has become “popular”. Therefore I enabled some more things to accomplish this:

• The ClamAV virus scanner provides inofficial databases by Sanesecurity to catch various sorts of spam. So there’s actually some kind of redundancy, as the virus scanner and the anti-spam filter are now sharing some responsibility. As not a single virus occured at my site in the recent months, this has now changed in some way.
• I installed the PDFInfo plugin from SARE and enabled it in /etc/spamassassin/init.pre.

Sure, as soon as we catch enough of that new PDF spam, spammers might change to some other document file format, such as DOC or RTF or even ODF, and we are forced to scan those attached documents for spam text or even for contained images that contain spam text, what we are already considering with FuzzyOCR. There must be some better way, actually.

However, I had to reduce the score of the Botnet plugin, as the default value of 5 points is way too high. Maybe I should add the plugin that checks the operating system such that only botnet clients using that Windows crap get high scores.

The fight continues.

Den folgenden Text habe ich vor einiger Zeit verfasst und heute zufällig wiederentdeckt. Ich hatte ihn in eine einfache Textdatei geschrieben, die vom 27. Februar 2003 datiert. Ich bin mir ziemlich sicher, dass er durch die Anekdote, die ich darin eingearbeitet hatte, motiviert war. Der Text ist als Gedankensammlung zu sehen und wurde hier nur minimal angepasst. Alle Rollenbezeichnungen sind geschlechtsneutral aufzufassen. (Schade, dass man das heutzutage immer dazusagen muss.)

Warum Mathematik? Was kann man damit überhaupt machen?

Als Mathematiker eröffnen sich einem dieselben Möglichkeiten wie für Physiker, Informatiker oder Elektrotechniker. Dies zeigen zahlreiche Stellenangebote entweder explizit oder in der Form, dass eine gewisse “oder vergleichbare” Qualifikation gewünscht sei. Die Welt ist nicht so trivial, dass Elektrotechniker unbedingt Bauteile zusammenlöten, Informatiker etwas auf der Tastatur tippen oder Mathematiker etwas an die Tafel kritzeln müssen. Die Stärke der Mathematiker ist, wissenschaftliche und technische Probleme formalisieren zu können, gegebene Abhängigkeiten zu bestimmen und eine potentielle Lösung wieder in verständliche Sprache übersetzt zu präsentieren. Während Ingenieure ein Problem eher als ein individuelles ansehen, haben Mathematiker und Physiker die Tendenz, die Allgemeinheit eines Problems zu erkennen und eine entsprechende allgemeine Lösung zu präsentieren, die zur Lösung des speziellen Problems angepasst werden kann.

Techniker sind Spezialisten. Innerhalb ihres Fachbereiches können sie schnell Antworten liefern. Ihnen wurden bei ihren Studien zahlreiche Kochrezepte beigebracht, die sie auch erfolgreich anwenden. Kommt ein Problem allerdings an den Randbereich eines Fachgebietes, stoßen sie auch an die Grenzen ihres Verständnisses. Ein Beispiel ist mir selbst widerfahren: Eine Gruppe von Ingenieuren rätselte, wie oft die Gläser erklingen, wenn eine Gesellschaft von 9 Personen miteinander anstößt. Ein Mathematiker kann ihnen die Lösung sofort sagen und sie ihnen nachvollziehbar erklären, er kennt nämlich die Allgemeinheit des Problems: Es handelt sich um die Frage, wieviele Möglichkeiten es gibt, eine fixe Anzahl von Objekten aus einer größeren Menge auszuwählen; dies ermittelt der Binomialkoeffizient. Die Ingenieure rechnen herum und präsentieren eine völlig falsche, mit Denkfehlern behaftete Lösung, was zeigt, dass ein grundsätzliches Verständnis der Problemstellung schlicht nicht vorgelegen ist. Zeigt man ihnen dann die formalisierte Version des Binomialkoeffizienten, ist sie in ihren Augen lediglich eine weitere, trockene Formel.

Mich wundert nicht, dass Informatiker, die aus der Technik kommen, bei uns nicht-technischen Mathematikern um Nachhilfe ansuchen, da sie erkennen, dass sie zwar zahlreiche Methoden erlernt haben, aber nicht verstehen, warum diese gerade so aussehen und in einer bestimmten Situation überhaupt angewandt werden können.

Mathematiker arbeiten vom Allgemeinen ins Spezielle. Techniker gehen eher tief in ausgewählte Teilgebiete eines Fachgebietes hinein, während Mathematiker es eher “der Breite nach” machen. Dafür müssen sie bei neuen Themen nicht mehr so tief hineinschneiden wie Techniker. Mathematiker lernen Neues schneller.

Absolventen dieses Studiums haben viele Möglichkeiten. Es gibt in der Tat so gut wie keine Arbeitslosen unter ihnen. Der Großteil, das sind etwa 38% (Quelle unbekannt, Anm.), kommt in der Informatik unter. Viele treibt es auch zu Versicherungen. Dort betreiben sie zwar nur mehr marginale Mathematik, aber das mathematische Denken haben sie verinnerlicht - die eigentlichen Inhalte des Studiums gehen nämlich verloren, diese werden nur in der Forschung gebraucht.