Slashdot Mirror

Oh, and you might have better work with Slash 0.3-3.7 than the stock versions that Rob puts out. Those versions are quite buggy. Oh, btw, there's also NNTP support for Slash.

I'm the one who put out newsd and the 0.3-3.7 tarball. I also seem to be making the most noise about adding functionality to Slash on slash-help. In fact, someone just offered to host a Slash CVS repository for me and I'm forking the code (see slash-help - especially the recent YA0.4W thread - for the gory details of why.)

(I've sent a copy of this reply to the original poster via email since I don't know if he's still reading this story or not.)

ASU also has a satellite getting launched at Vandenburg AFB. You can check out the website at http://www.eas.asu.edu/~nasasg/a susat/asusat1.html. It's got lots of info on the design and stuff. The ground control software is written in Java and the development platform for that is Linux. Pretty cool project.

This BBC article yesterday about the discovery of yet another Kuiper Belt object by Pioneer, mentions at the bottom:

Earlier this year, scientists were puzzled by what was described as a mysterious force acting on the probe. It led to speculation that there was something wrong in our understanding of the force of gravity.

Eventually the effect was tracked down to the probe itself, which was unexpectedly pushing itself in one particular direction.

I expect this new theory will also be dispelled by minor impacts, leaking remainders of fuel, and the fact that space isn't a true vaccuum. I'd be delighted to be proved wrong, of course.

--

Just try the 28 hour day!

Well, um, what about those of us who took the Slashdot code and did something with it, namely fixing the bugs, making it so it runs pretty much out of the tarball, and writing a FAQ on how to use it (http://womensmusicnews.com/null/, plus manning a mailing list (http://projects.is.asu.edu/m ailman/listinfo/slash-help so Rob had to deal with fewer lamers and got some code-level feedback? Don't we get something more out of this? Yeah I'm way biased here, but I'm also totally broke (what savings account? what money?) so I'm looking out for myself too. ;)

Of course, Rob's and Andover's response is "no comment" (couldn't think of a way to use C or Perl-type comments in an expression of "no comment").

One thing I find sorely lacking in many books on algorithms is any discussion of why you would select one over another

Do you mean GA versus say a Newton search method? GA is sometimes referred to as a method of last resort. This may be unfair, because many practical problems are not mathematically "nice". I am just getting into GA and I have very complicated simulations underlying my objective functions. We previously computed derivatives for these; it was a huge effort both computationally and for the programmer. One thing that I like about GA is that wrapping the optimizer around an arbitrarily complex objective function is really easy. Also, the parallelism is really good ("embarassing"), especially for distributed computing with message-passing (think beowulf).

For me, the bad thing is that convergence isn't nice and quadratic like some derivative based methods out there. On the other hand, quadratic convergence generally works only near the optimum and derivative based optimizers really only find local minimums (no guarantees about the optimum being globally optimal). Derivative based methods can blow up if you pick a bad guess objective too. Perhaps a good strategy is a combination -- use GA to get into the neighborhood of the global optimum and then use derivative based methods to find it.

I should stress that all this is for my particular application (groundwater). YMMV. Others with different objectives living in differently constrained control spaces will have different experiences. Also, to be fair I should point out that programs like ADIFOR make derivative computations easy to program.

Some helpful optimization links:

Decision Tree for Optimization Software

GA Archives

MOSIX does not work with a simple kernel patch, it requires a whole new kernel. For example, MOSIX manages and runs two processes for each UNIX-style process. Furthermore, for a program to run in-parallel, it has to be written as a parallel program. Look at Chime for a better solution.

>Post new version of slashdot code

Er... http://projects.is.asu.edu/m ailman/listinfo/slash-help would be a good place to start if you want to even bring up Slash.

There's a little link that says code on the page somewhere. It leads to Slash 0.2 and 0.3-pre. Please report to the eyedoctor. Oh yeah. Please report to the Slash-help mailing list too for information.

Ooooh a -1, remind me to redefine the Slashdot Quotient to account for that.

BOFH mode continues to be on.
Slash-help mailing list, Scary Devil Monastery for recovery after installing Slash. Remind me to put in the FAQ on how to turn moderation OFF too. Darn hypocrits.

Sigh, this is going to ramble, but actually Disney was in a bit of trouble a year or so ago when the copyright or whatever on Mickey Mouse expired and he would suddenly be in the public domain, and anyone could make profits on him. I forget how they solved the issue,

For more information and links (from a law professor opposed to the extension), follow this link.

There's more than one pulsar system. PSR1829-10 was retracted. PSR1257+12 most definately was not and is considdered confirmed with multiple planets. (as much as any can be considdered confirmed. No extrasolar planets have been directly imaged)
See http://www.public.asu.edu/~sciref/exo plnt.htm

Even a spell checker doesn't solve all your problems, as this classic poem demonstrates:

Owed Two the Spell Checker