Slashdot Mirror

This is the kind of thing that inspires me to do a PhD in computer science. I think all the techniques in the video were covered at some level in lectures at university (mostly this course, IIRC) but seeing so much of it working in sequence with real photographs is impressive. (Of course, I saw impressive stuff at university. And this example is cool, whereas the most impressive computer vision work at Imperial College is medical, and I was too squeamish and usually distracted by the blood and skeletons.)

Alternatively, I need to find a job that stretches my ability, rather than an easy/comfortable one.

And your statements about reliability? In what sense can a logic circuit be "guaranteed" free of defects? Did Intel know about this method of quality assurance back when they were designing the Pentium? It seems to me that simple logic circuits can be guaranteed free of defects because the human mind can readily model the whole system and intuitively decide it is correct. When the system is complex, that is no longer true.

There is some progress being made towards "guaranteeing" the correctness of circuits, such as:

this. Centaur Technologies (VIA) uses theorem proving tools to guarantee the correctness of parts of the VIA Nano processor. I'm sure with a little digging more references to this sort of thing can be found.

Intel appears to be actively working in the area of formal verification also, e.g. this - although this doesn't directly deal with low level circuits.

So, it is possible to guarantee some correctness, although I suspect it rests on the correctness of the theorem prover you're using also.

Actually Photosynth started off as a PhD project called Photo Tourism. You can download the source code of Bundler (GPL license) as well. The idea of using geo-tagged photos to create a 3D view of the world is really cool. However there are various challenges: occlusion, moving objects (people, foliage, ...), changing illumination, different cameras, cameras with distortion. The software needs to be robust and discard those "outliers".

SceneLib is a software for simultaneous localisation and mapping (SLAM) which essentially is a similar problem. However here the assumption is that the camera is always the same.

Creating a panorama with and Enblend on the other hand only allows pictures taken from a single view point.

Much of what you describe (the depression bits) applies to my time at Imperial College London (an MIT of Europe?).

Firstly, no one graduated early, and since the debt is manageable or non-existent (depending on circumstances/citizenship) for EU students it would need to get really bad to feel you were wasting money. Of the people that left -- I'm guessing about 10% -- most gave up because it wasn't for them, and as a consequence failed the exams at the end of the first year (and chose not to retake the year).

In my final year I lived in a flat with two guys, one became quite depressed in about October (I don't know why), the other was hardly in the flat and was really dull -- a stereotypical nerd. In retrospect, I was stupid to choose to live with them, and should have gone into a thousand pounds more into debt to live with my friends (who wanted a nicer flat than I thought I could afford). The depression was contagious, my flatmate's constant bitching about the university and his complete lack of motivation ended up getting me down and definitely affected my grades.

What really helped was socialising with students from other universities, and to a lesser extent younger students from Imperial. I'd made a few friends with students from other London universities and found myself spending a lot more time with them to escape the stress (of my final year friends at Imperial) or depression (of my flatmate). So, during my "free time" I was fine, I was still motivated to go out clubbing/partying, but I'd return to my flat and *flop* not get anything done. I eventually realised I was depressed and found the staff to be sympathetic, although I hadn't turned to excessive drinking anyway (the doctor did ask if I was using drugs).

(Should mention that it's obviously not only my flatmate's fault, I was susceptible to the depression. But if I'd been living in a happier home my mood would have been much better for much more of the time.)

You should read about Igor Aleksander's WISARD project although you might be better off reading one of his papers rather than spending on a book http://www.iis.ee.ic.ac.uk/aleksander/publication.html

I don't know about that. Certainly at Imperial College London (where I happen to study), the courses run by the Department of Computing are varied enough that a lot of the skills necessary of a games programmer can be gained (graphics, AI, computer architecture, etc).

Although it might fair to say Imperial has an unusually strong link with the games industry, e.g. the Games and Media Event and EA has run an event on campus in the past.

The article is available at the author's website.

As far as I can tell, the original result provided a conformal map from a disk onto a polygon. Prof. Crowdy extended this result to provide a map from a disk with circular holes poked in it onto a domain with polygonal holes. Why is it useful? I am sure someone in the applied camp would know.

The article is available at the author's website.

As far as I can tell, the original result provided a conformal map from a disk onto a polygon. Prof. Crowdy extended this result to provide a map from a disk with circular holes poked in it onto a domain with polygonal holes. Why is it useful? I am sure someone in the applied camp would know.

This is not true.

Yes, it is.

This is not true.

Yes, it is.

Here is the old theoretical paperOne path to acoustic cloaking, New Journal of Physics, v. 9, 45, 2007. [pdf reprint] The Science Daily article is just a reprint of the Duke press release. Steven A. Cummer seems to provide PDF "reprints" of all his papers but the new one isn't in that list. Nor can it be found on David Smith' page, David Schurig's old Duke page, or his new NC State page, Sir John Pendry's page, or Anthony Starr's page.

Coal is not that cheap, and pollutes more than anything else

We have a few machines like Dinorwig which feed peak demand from baseload generation.

We have a lot of windmills that are politically correct, but sited where they disfigure the environment, and generate no electicity at all, as far as I can see.

A map of coverage in Europe from *1997* has about 50% GSM coverage. A current map has pretty much 100% coverage.

The CIA factbook says the USA has 233M subscribers (2006), 301M people = 77%
The EU has 466M (2005), 490M people = 95%
(OK, your link is newer. But look at the subscriber ratios in the boxes on the map for the UK, Germany, Sweden etc -- above 100%! My 85 year old grandparents do have mobile phones ;-)

Perhaps you didn't get much out of neural networks, but my PhD thesis was on the similarities between a kohonen network and relaxation-labelling equations. Part of it is up on my blog (I haven't actually got as far as that bit yet, but the groundwork is there).

A neural network (well, anything more complex than the single-layer perceptron anyway) is an arbitrary classifier. I'm curious as to why other methods are "much better". Unless you do an exhaustive search of the feature-space, all classifier methods are subject to the same limitations - local maxima/minima (depending on the algorithm), noise effects, and data dependencies. All of the various algorithms have strengths and weaknesses - in pattern recognition (my field) NN's are pretty darn good actually.

It's also a bit odd to just say 'neural networks' - there are many many variants of network, from Kohonen nets through multi-layer perceptrons, but focussing on the most common (MLP's), there's a huge amount of variation (Radial-basis function networks, real/imaginary space networks, hyperbolic tangent networks, bulk-synchronous parallel error correction networks, error-diffusion networks to name some off the top of my head), and many ways of training all these (back-prop, quick-prop, hyper-prop, batch-error-update, etc. etc.) I guess my point is that you're tarring a large branch of classification science with a very broad brush, at least IMHO.

Not to mention that this is all the single-network stuff. It gets especially interesting when you start modelling networks of networks, and using secondary feature-spaces rather than primary (direct from the image) features. Another part of my thesis was these "context" features - so you can extract a region of interest, determine the features to use to characterise that region, do the same thing for surrounding regions, and present a (primary) network with the primary region features while simultaneously(*) presenting other (secondary) networks with the features for these surrounding regions and feeding the secondary network results in at the same time as the primary network gets its raw feature data. This is a similar concept (if different implementation) to the eye's centre-surround pattern, and works very well.

If you work through the maths, there's no real difference between a large network and a network of networks, but the training-time is significantly less (and the fitness landscape is smoother), so in practice the results are better, even if in theory they ought to be the same. I was using techniques like these almost 20 years ago, and still (very successfully, I might add) use neural networks today. If it's a fad, it's a relatively long-running one.

Simon.

(*) In practice, you time-offset the secondary network processing from the primary network, so the results of the secondary networks are available when the primary network runs. Since we still run primarily-serial computers, the parallelism isn't there to run all of these simultaneously. This is just an implementation detail though...

http://www.doc.ic.ac.uk/~matti/ise2grp/energystorage_report/node5.html

Renewables have a big problem - solar and wind are unreliable. It is simply inconceivable that electricity supplies can fail when it is cloudy and/or not windy. If you want to build renewables you need to also build gas-fired plants (usually, other possibilities exist) as a backup. That makes things look very expensive

The site's gone down, so here's a copy of the torrent file:

rms-talk.ogg.torrent

I didn't get the Q&A torrent.

It's hard to make a midi-tower sexy.

It's hard to make a midi-tower sexy.

Well, yeah, that is a problem. FDR is free for academic use, IIRC. But they charge for commercial use. Although, as the Formal Systems Europe folks point out, exactly how much debugging time do they need to save to justify the cost of FDR.

If FDR is too pricey for you (perhaps you're an open source developer), there are a few other options. In the free category is LTSA. LTSA accepts a language called FSP, which is similar (but not identical) to CSP. The tool itself is not as powerful as FDR from what I can tell, can potentially still very useful. And there's an Eclipse plugin :-) In the open source category is SPIN. SPIN's process modeling language, Promela, is more C-like than CSP-like, but does have CSP-like semantics. The tool itself is free, open source, very powerful, and has an active community of users. Definitely worth a try if you're looking for something capable of industrial-strength model-checking.

Look at CO2 (a gas) and SiO2 (silica, a solid).

That's a misconception; the sort of silicon-based life that we're talking about are not precisely the same as carbon chains. In carbon chains, you typically have C-C-C-C-C... etc. In the equivalent silicon (actually silicone) molecule, you have Si-O-Si-O-Si-O-Si.... etc. Si-Si-Si... etc doesn't chain well, but Si-O-Si-O.. chains indefinitely. Compare a hydrocarbon-based lubricant with a silicone-based one, hydrocarbon solids (plastics) with silicone ones, etc. There's been a lot more research on the former so far; the latter can likewise be functionalized.

A few differences in the chemistry:

1) C-C-C-C-C... chains can freely rotate, while Si-O-Si-O... chains need a specific "joint" to do so.
2) Carbon more readily double and triple bonds, although removing Os from the Si-O chain can create similar (but not equivalent) effects.

There are all sorts of biologically interesting silicon compounds. The silicon equivalent of methane is silane. It's even more flammable than methane; it's hypergolic with our atmosphere (burns on contact). Its giving up of its hydrogen could be seen as equivalent to ATP and its phosphorus. Longer "silanes" scale like longer hydrocarbons -- their vapor pressure decreases the longer they get (silanes with 2-3 silicons make for good wood sealants). Zeolites are silicates (your typical silicon solids that you were picturing) but with various metal ions interspersed with them; they're excellent, highly selective catalysts. Probably the most biologically interesting (to me, at least) are silanols, which exist naturally in Earth's oceans (and probably predated life), and can form all sorts of catalytic groups, membranes, etc.

Of course there is a replacement for Outlook and Exchange! It's called sendmail and it's part of every unix-like system. You install an MTA (either the original sendmail or a compatible replacement) and a POP3 server on a machine (an old desktop is fine), configure your firewall to route incoming traffic on port 25 to that machine, log into your DNS control panel, and set its internet hostname as the MX for your domain. Then you run a normal mail client on each desktop. Specify your mail server's inside IP address as the SMTP and POP3 server in your mail client, and away you go.

Since when does chocolate have a significant amount of caffeiene? Maybe you're confusing it with Theobromine? There are some conflicting accounts regarding caffeine in chocolate, but I'm inclined to believe the info on that page mostly because trying to drink coffee (even when it's been diluted to hell and back with milk and syrups from the local barista) starts to give me a headache after just a couple of ounces, whereas eating a large amount of dark chocolate at once makes me feel relaxed and kind of floaty (probably more due to the subsequent buildup of Anandamide than the Theobromine itself.)

I see you are from NZ, I visited there last year, it's a beautiful country....

I've always found it easier to recognise when someone's stoned than coked up, but the thing is, when stoned, people tend to compensate and drive slower and put more attention into it, often to the extent that they're actually more careful, safer drivers while stoned, then while straight. But then the effects of green are somewhat different, the biggest danger is that you'd forget where you're going and end up somewhere else *lol*

"and most who use coke don't do it while drinking (it would be a waste)"

That's completely untrue. "When there is alcohol and cocaine present in the blood stream, as is usually the case, the two form an even more potent stimulant - cocaethylene, an ethyl homologue of cocaine" (see 1, 2)

So not only do I not believe claims that facial recognition "is coming soon". I don't believe claims that it's "already here". I don't even believe the demos done right in front of my eyes, because we all know what demos are like. I'll believe it when a credible organisation trusts their money or security to it for an extended period.

Another nice visual SLAM (Simultaneous Localisation And Mapping) page is Robert Sim's http://www.cs.ubc.ca/~simra/index.html , where he uses multi-camera vision to produce a much denser map.

A Brief History of Quantum Computing contradicts all your quantum-computing assertions: "In effect, a calculation performed on the register is a calculation performed on every possible value that register can represent." That's in its description of Shor's algorithm, which also contradicts your feedback-driven characterization, saying it produces very-likely factors and succeeds by simply retrying until one of its answers works.

That link also describes Grover's algorithm, cutting brute-force search from O(N) to O(N^0.5). That alone is enough to put AES-128 in range of today's horsepower (but not enough to reach AES-256). Maybe it's provably impossible to reduce symmetric decryption to less than linear search, I don't know.

The number of plausible-but-wrong decryptions of a cipher block you know the plaintext for is zero, so unicity distance is only trivially relevant; if you insist, we can say it's exactly the cipher block size regardless of the plaintext language in this specific case.

I totally agree with this article. The issue is not only that we don't focus enough on the rest of the cycle, but also that we have no understanding of it.

The best work in the field is the so called "FEAST hypothesis" http://www.doc.ic.ac.uk/~mml/feast/. What they say is that systems evolve along an inverse square rule. But there are catches. The big issue in systems is not growing size, but growing complexity. Simple code is easy to maintain, complex code is a nightmare. The biggest catch though is in the data. I have later data for one of the systems studied by the FEAST project. It grew from 3.1 mloc to 3.9 mloc in the projects study period, along an inverse square form. If the system had continued to grow at that trend it would be 4.5 (ish) mloc today.

In fact it's 13mloc and it's become the nexus of an ecosystem of 400 systems connected to it via XML middle ware.

We know squat.

*Sigh* Every year or two for the past ten years, I get into an argument with some European professor about the qualifications of American graduated students, Usually on their knowlege of engineering of computer programs.

Thanks to this discussion, this year, I am going to have a much harder time defending them, I think.

Safety critical systems is a branch of computer science, routinely taught in the better European universities. The book I first learned it from was Safety Critical Computer Systems http://www.eng.warwick.ac.uk/~neil/safebook.htm

The field really got impetus after the Therac-25 failure http://courses.cs.vt.edu/~cs3604/lib/Therac_25/The rac_1.html
where a group of people got radiated to death because of a minor (in a computer sense) error in some code, thus adding new meaning to the term "execution error".

Its a vast field (google "safety critical") with large numbers of interesting published papers and few good books. I really can't remmend a current book, I haven't seen one I liked in years. I don't think you can learn it fast enough to be useful.

You also need to know a fair amount of higher math to be really competant, or at least even understand whats going on... most of the true experts in the field apparently regard english as a required second language (Math being the first language, not that I can blame them, but often it is overused)

The field seems to have suffered since the recession. and fragmented, but there is a good starting point at http://vl.fmnet.info/safety/

It also seems to be rapidly migrating to India as well, because of the resistance of American "cowboy" programmers. This time, it is possible the game of "cowboy and indians" may end up with the Indian's winning, inasmuch as the techniques are going to be essential to the new multicore programming models. (I heard a rumor Herb Sutter is investigating that, but thats just a rumor. If so, however, he would be the person to talk to about safety critical C/C++)

One of the techniques used in the field is formal verification. McGee and Kramer are coming out with a second edition of thier incredible book, Concurrency, http://www.doc.ic.ac.uk/~jnm/ early this year (last time I emailed them). The book is an gentle introduction to the field of formal verification and model checking, among other things. There are other books (a new one came out on the Spin Model checker, for example) but this is by far the most penetrable.

Much more interesting is the use of model checkers behind UML or BPEL/SOA tools.
Most of the really interesting stuff is still behind university walls, but tools should be appearing soon. I am trying to develop an open source grid based one but it's been slow going due to committments and resources.

Have you heard of Java subroutines? They are a clever mechanism designed to reduce code bloat from nested finally clauses. To make a long story short, subroutines account for more than half the length of the formal correctness proof for Java, and complicate the implemention of the compiler and runtime system substantially. The punchline is, changing the name from "Java" to "Oak" would, on average, save 13 times as much space as adding subroutines. Well designed with a theoretical basis in mind? Decide for yourself.

As part of an empirical study, I've used CCCC , to measure the software evolution of software (how metrics for SW architecture evolves over time). This tool supports McCabe's cyclomatic complecity, LOC, Henry-Kafura information flow metric, module coupling etc. The purpose can be to detect software decay, architecural mismatch etc. However, to study software evolution, one needs access to all the versions of the software, which this site does not provide (yet).

Lehman did the first studies on software evolution, proposing several laws for the field.

http://www.newton.dep.anl.gov/askasci/chem99/chem9 9562.htm

That's from 5 sec of searching, but it is true. HCN binds to the heme quite strongly.

You of course are right in that cyanide kills by interfering in the electron transfer process as HCN binds to cytochrome a3 in mitochondria. HCN does bind to the heme in hemoglobin, too, just not as well as CO.

To put it in modern terms, they're CAV devices.

Actually, that's why Apple are moving from PPC to x86, but don't tell anyone ;-)

This is actually part of the very extensive 'Box Mine', which was excavated to build the houses in the nearby city of Bath. I visited 'box mine' in 2004 with a caving club, there's a locked MoD door somewhere amongst the labyrinthe of tunnels excavated over the last 150yrs.

http://www.union.ic.ac.uk/rcc/caving/photo_archive /trips/2004-10-17%20-%20mendips%20-%20jarvist/diri ndex.html

boxen
 
/bok'sn/ (By analogy with VAXen) A fanciful plural of box often encountered in the phrase "Unix boxen", used to describe commodity Unix hardware. The connotation is that any two Unix boxen are interchangeable.

Well, as I understand it, there were some earlier efforts, such as Speedcoding by Backus himself, so people had some sense that you could program at an abstraction above the machine level. From what I've read, things like Speedcoding weren't fast, and so speed was indeed viewed as a big hurdle. That said, my impression is that languages like FORTRAN were much more comprehensive and ambitious than earlier efforts, so your implication that people viewed it as magic (ahem, I meant AI) may be quite right.

http://wombat.doc.ic.ac.uk/foldoc/foldoc.cgi?query =scripting&action=Search -- Nothing in there implies a "scriping language" isn't a programming language. Ruby and python can be used for scripts, are they not programming languages now?

Yeah that's what I want, all my applications to be server side web-based. That way I can't stop them when they "call home" and report back things like what I'm searching for on the net, the names of the files I'm opening. And I can't stop them from a hacker switching out a DLL on the server side suddenly corrupting or infecting my data. AV and firewalls become useless at that point, and the way modern apps try agressively to monitor what you do and call home, I'm not comfortable with losing the ability to control them.

You have no control over software right now. Unless you build it from source yourself, and inspect the source, and are capable of editing it if you find something you don't like. If you're running a window system, it could call home until the cows come home, inform everyone of your every action, hold your documents hostage, have dll's swapped out from under you, and many other things. Moving to the web will DECREASE these risks, not increase them, because securing a desktop from malicious third parties is harder than securing a server from the same.

In any case, server-side will have backups of your work, so if it gets corrupted, there will be something to restore it from. If you install a new software patch on your home system, and it corrupts your files, and you were lazy with the backups, as most people are, your files are gone.

The control you think you have is an illusion. Backdoors have been built into widespread systems WITH source availability without being found out. You just don't know if you can trust your system.

Though I must admit, any web service that tries to keep me from getting to my own files, I will not use. In the end I trust nothing and no one, so diversification of where and how I keep my data is the key.

What makes you think that banks have a clue about computer security? There exists evidence to the contrary.

That is so NOT the acronym. Actually, TWAIN isn't even an acronym at all. See this to be humbled a bit.

No, jonadab was correct. foldoc link

And the chips would be cool also...

A little rose of humour blooms in the darkest depths of the Pentagon's labyrinths!

What about Shub-Internet?

According to this the radiation is spread fairly evenly across the entire skull. This study had interesting numbers, but the methodology seemed odd (they apparently stuck the antenna 4cm in front of the eye, instead of at the ear). If the first document is correct, that would explain why FDA uses SAR as a metric: it's invariant throughout the head. It would also put the numbers back in the "concern" region.

I'm not going to stop using my cell, but I'm going to watch the research with interest.

Regards,
Jeff Cagle

Automatic buffer overflow protection only covers the straightforward buffer overflow problems, i.e. array index overflows. In the case of more complex pointer arithmetic, where most of these problems occur, automatic protection is not possible (at least not without losing the option of pointer arithmetic).

Actually, automatic checking is very much possible, and has been for years. For example, Bounds checking gcc (that website is down right now, so try my page on the subject). That was written in 1994, and there are newer systems available now which don't have such a serious performance penalty.

The real solution is to stop writing critical code in C. Other languages provide bounds checking, and are faster and safer than C: for example OCaml which I prefer nowadays.

Rich.

This is not the case; Actually the information needed to describe a system of let's say N atoms grows exponentially. So you don't need that much atoms, or Josephson junction circuits at all... see http://www.doc.ic.ac.uk/~nd/surprise_97/journal/vo l4/spb3/ for a short introduction.

This is called "making it up as you go along"

Are you saying this because you don't know what Huffman coding is? It's indisputable that more frequent letters in english get shorter codes, just like Huffman coding. So using morse code for, say, Czech, would probably do a lot worse.

I'm a bit rusty from my thermodynamic courses, but I'll try: it's in transition state where both liquid and solid exists. So up to 6.5 it's solid, from 6.5 to 7.6 it's in transition and 7.5 and above its liquid. Can anyone refute or back this up? I'm not a chemical engineer, I only took courses way back when...
This might help. Or This (pdf) or This (pdf)