tag:blogger.com,1999:blog-67983712512644291612012-04-16T04:54:42.449+01:00First, computer science is not about computers. Second, computer science is not a science.Manuelnoreply@blogger.comBlogger13125tag:blogger.com,1999:blog-6798371251264429161.post-12621780487277920702009-04-15T12:53:00.004+01:002009-04-15T12:58:33.262+01:002009-04-15T12:58:33.262+01:00

The central problem of reinforcement learning is to fit the parameters of an agent's policy in order to make the agent chose "good" decisions according to some unknown objective. Let's consider a very simple case. We have an MDP which is basically stateless. So to say, all the world is doing is to map the actions of an agent directly to a reward. Consider this function to be a mixture of two

Justinhttp://www.blogger.com/profile/04972102120598633110noreply@blogger.com1http://www.bitwiese.de/2009/04/when-stochastic-search-beats-direct.htmltag:blogger.com,1999:blog-6798371251264429161.post-46150013019681815482009-02-13T09:01:00.003Z2009-02-13T09:20:38.486Z2009-02-13T09:20:38.486Z

Did you ever wish your program had a pause button? I did so regularly. I regularly have to do lengthy calculations, for example for evaluating some heuristics for an optimization problem such as Maximum Independent Set or Matchings. A typical program could look like this: for all benchmark graphs G: for all algorithm variants A: for k in [2, 4, 8, ..., 128]: Execute A on G with the

Manuelnoreply@blogger.com0http://www.bitwiese.de/2009/02/bash-job-management-essentials.htmltag:blogger.com,1999:blog-6798371251264429161.post-33966257330800351532008-12-16T18:15:00.008Z2008-12-16T18:33:57.097Z2008-12-16T18:33:57.097Z

This is mainly here as a reminder to myself and anyone else who might fall into this. The scenario is as follows: I want to compile a C Python extension that uses OpenGL on Mac Os X. My setup.py file looks something like this: #!/usr/bin/env python2.5 from distutils.core import setup, Extension module1 = Extension('foo', sources=['foo.c'], extra_link_args=['-framework OpenGL']) setup( name='

Manuelnoreply@blogger.com3http://www.bitwiese.de/2008/12/pitfall-with-distutils-opengl-on-mac-os.htmltag:blogger.com,1999:blog-6798371251264429161.post-51606909638750436872008-02-13T10:26:00.009Z2008-02-13T12:54:24.942Z2008-02-13T12:54:24.942Z

Python is slow in terms of execution speed. Period. It is fast enough for most tasks out there, but when it comes to number crunching or manipulation of lots of objects, you can hear the interpreter grind its teeth. Python advocates have a lot to say on how to handle these situations. The most common is "you can always rewrite that particular part of code in C". There are other recommendations:

Justinhttp://www.blogger.com/profile/04972102120598633110noreply@blogger.com5http://www.bitwiese.de/2008/02/gaining-speed-with-weave.htmltag:blogger.com,1999:blog-6798371251264429161.post-26081044581768707872007-09-30T22:09:00.000+01:002007-09-30T22:35:49.776+01:002007-09-30T22:35:49.776+01:00

C++ is a great programming language. It allows you to mix the most important programming paradigms like object orientation as well as generic, functional and structured programming. Another nice aspect is that the compilers are efficient and with some black template magick, you get some of these things for almost free at run time. However, one of the greatest features of C++ can hurt you:

Manuelnoreply@blogger.com0http://www.bitwiese.de/2007/09/c-beware-what-lurks-beneath-surface.htmltag:blogger.com,1999:blog-6798371251264429161.post-33053098629934155142007-09-26T15:46:00.000+01:002007-09-27T08:24:52.015+01:002007-09-27T08:24:52.015+01:00

Note that I made a grave error of thinking before writing this article: I forgot the copy-on-write page sharing of modern Unices. I added two paragraphs to this article that should clarify the point. Thanks for your comment, Alex. Please note that I do not want to critizize Jason here. I am simply picking up this point from his presentation as a motivation of a topic that I wanted to write about

Manuelnoreply@blogger.com11http://www.bitwiese.de/2007/09/on-processes-and-threads.htmltag:blogger.com,1999:blog-6798371251264429161.post-33823778325963648862007-08-07T22:08:00.003+01:002011-11-21T12:53:10.271Z2011-11-21T12:53:10.271Z

Acknowledgement (2011-11-21): You should search Google for your blog titles before hitting "publish". When I wrote this article some years back, "hitting the memory wall" was just a term I had read some time earlier and internalized for myself as a standing term, like folklore. This was wrong. There is a paper from 1995 by WA Wulf and SA McKee with the same title: Hitting the memory wall that I

Manuelnoreply@blogger.com8http://www.bitwiese.de/2007/08/hitting-memory-wall.htmltag:blogger.com,1999:blog-6798371251264429161.post-15423780301782809782007-06-03T23:01:00.000+01:002007-09-26T15:41:19.338+01:002007-09-26T15:41:19.338+01:00

Acknowledgements: This article is based on the work of Peter Sanders and Roman Dementiev. You can find the full source code presented here in their STXXL library (STL library for large data structures), Peter Sanders' paper and the accompanying source code. Warning: If you are of sensitive nature either regarding computer science theory highly practical program optimization then I

Manuelnoreply@blogger.com10http://www.bitwiese.de/2007/06/highly-efficient-4-way-merging.htmltag:blogger.com,1999:blog-6798371251264429161.post-35361501949508770242007-05-17T15:46:00.000+01:002007-05-25T12:08:02.387+01:002007-05-25T12:08:02.387+01:00

The language war will always go on and I consider myself as someone who has to promote the Python programming language no matter what happens (even though I know that there is no silver bullet.) Thus, I do have to use every opportunity to show the world how easy it is to connect formalisms of high scientific importance with not-so-important-but-still-cool language features. Whatever. Markov

Justinhttp://www.blogger.com/profile/04972102120598633110noreply@blogger.com1http://www.bitwiese.de/2007/05/simulation-of-markov-decision-processes.htmltag:blogger.com,1999:blog-6798371251264429161.post-90003694984531223762007-04-19T22:24:00.000+01:002007-04-19T22:51:57.113+01:002007-04-19T22:51:57.113+01:00

TextMate's TODO bundle lets you keep track of the TODOs in your projects. You can mark a line to show up in the TODO list with FIXME, TODO or CHANGED. The following two images show a text file with these keywords and the resulting TODO list. The only problem is that if you keep a copy of Rails in vendor/, then the TODO plugin will also show the items in Rails (or any other external libraries

Manuelnoreply@blogger.com1http://www.bitwiese.de/2007/04/ignoring-folders-with-todo-bundle-in.htmltag:blogger.com,1999:blog-6798371251264429161.post-17721751432210373902007-04-11T18:53:00.000+01:002007-04-11T19:30:10.075+01:002007-04-11T19:30:10.075+01:00

Sometimes, floating point numbers are the thing that futile hours of programming are made of. Whether they come as 32, 64 or any other precision you want, they can truly give you unnecessary and unexpected work if you do not use them correctly. The problem I stumbled over yesterday was the problem of rounding errors. Consider the following C snippet double pi =

Manuelnoreply@blogger.com4http://www.bitwiese.de/2007/04/joys-of-floating-point-numbers.htmltag:blogger.com,1999:blog-6798371251264429161.post-87687442720225770892007-04-07T15:34:00.000+01:002007-04-11T18:55:13.027+01:002007-04-11T18:55:13.027+01:00

SQLite is great. No, really, I do mean it. If you ever need to work on structured data and editing it from the outside is a must, then SQLite is worth checking out and might be a superior choice to XML and will be a superior choice to your own proprietary format. I cannot speak about performance - there are some outdated claims that SQLite is faster than postgres and mysql for the most common

Justinhttp://www.blogger.com/profile/04972102120598633110noreply@blogger.com1http://www.bitwiese.de/2007/04/sqlite-trapdoor.htmltag:blogger.com,1999:blog-6798371251264429161.post-567032794957710492007-04-06T20:38:00.000+01:002007-04-06T20:47:20.506+01:002007-04-06T20:47:20.506+01:00

Obviously, this is the first entry of the shining new Bitwiese Weblog. Feel warmly welcomed by these few bytes which propably travelled kilometre over kilometre from here to Blogger's servers and back again to you. This weblog is brought to you by Justin and me, Manuel. We hope to publish information related to the following programmingcomputers in the widest sensecomputer science (prepare for

Manuelnoreply@blogger.com1http://www.bitwiese.de/2007/04/first-post.html