Quite the opposite, actually. There is evidence that in cases of cardiac arrest (where the body is generally healthy aside from the fact that the heart has stopped), slow revival can allow for a higher success rate after longer periods without oxygen, because the cells themselves only die hours after cessation of blood flow. If you read to page 2 of that link, you see that induced hypothermia is sometimes used precisely because it does help slow the process of cell death which follows clinical death. Granted, as far as I'm aware, we don't know that cryogenic freezing would suspend all of such processes, but the state of research in this area is much farther along than you seem to think.

I was under the impression that the important part of PD was that the dominating strategy is not globally optimal -- hence the "dilemma": if you both choose the dominating strategy, you both do worse than if you both choose the vulnerable strategy. The idea of a one-shot is, in my opinion, the most artificial aspect of PD, as in the real world even a one-shot PD is merely one of a series of PD games played among one large player pool, and thus I believe many people consider that "what goes around comes around" even when faced with a supposedly one-shot PD.

Seriously? A safe bet? Did you even read the rest of my comment? Wow. I... I can't even form a coherent response. Excellent troll, sir.

I think we need to define terms here. GP was talking about dogs bred for aggression to the point of bring mentally ill psychopaths. While I do see that the page you link says "They are loyal to their duty in protecting their flock and family and will stand by and defend through any circumstance. They are an extremely aggressive animal and need very experienced owners.", I think that is aggression in the sense that they will be proactive in defending their flock and family, even to the point of attacking a stranger if the dog feels the person is a threat. (As opposed to a less aggressive dog which might bark and growl without attacking, or on the extreme other end perhaps run away or greet the stranger hoping for affection.) Note that the same page says "vicious temperaments are considered serious faults for the breed." So I will acknowledge that there are breeds which are specifically bred for aggression, but I will stand by my opinion that there is no breed which is "mentally ill" or "vicious" as a whole, and especially that banning breeds of dogs is not a viable approach. If you ban dogs based on breed rather than individual training and/or behavior, you will still fail to address the real problem, which is that there are people who desire having dangerous dogs. Those people will find a way to make a golden retriever vicious if necessary, if you ban all the more aggressive breeds, and then you're back to square 1.

Like which ones? Rottwielers? I have only known 2 personally, and they were both incredibly sweet dogs. Sure, that's anecdotal evidence, but since you're making sweeping generalizations, I think it's better than nothing. If you want to convince me all Rottwielers are mentally ill phsycopaths, you're going to need some hard evidence.

I think you're wrong. They've been bred to have dominant personalities, and to be mistrustful of strangers as opposed to those they recognize as part of their pack. Neither of those things translates to "aggressive" in my book, much less "violent without provocation" which is what you seem to be suggesting.

Irrelevant to a discussion of their personalities. You might as well say that because body builders have spent years of their lives training to be bigger and stronger, and thus are more dangerous should one choose to attack you, they are therefore more dangerous people. Should we say people shouldn't be allow to body build? Or what about train in martial arts? They work to lessen their pain response in their skin, train to have stronger, faster muscles, and harder bones. Does that imply they are more likely to attack you without reason? Often, the opposite is true -- trained martial artists will be more confident and in control of their abilities, and are thus less likely to use excessive force accidentally through error or panic. (The same is true of properly trained dogs, regardless of temperament.) If a person attacks another unreasonably, we punish them. We should do the same (conceptually) for dogs, punishing the owner responsible for the situation if a dog attacks a person unreasonably. If it appears the dog is too dangerous to be safely rehabilitated, euthanize the dog. Banning certain breeds is still an inadequate and ineffective way to address the root issue. It's like saying that banning hollow-tip bullets will prevent shootings, because they are designed to do damage. At very best, it will change the situation such that if someone gets shot, they will get hurt less. But you know what? They real problem is that someone shot someone else, not the kind of bullet involved. Reread my comment -- banning certain breeds due to their physical characteristics will not reduce the number of "vicious" dogs in the world in any significant way, just like banning hollow-point bullets won't reduce the number of wackos with loaded guns.

I disagree. Some dogs are bred to have physical advantages in a fight. That isn't what makes a dog dangerous, it is their behavior and the ability of their owner to keep them under control in stressful circumstances.

... when properly raised and trained, yes. When improperly raised, most dogs make bad pets; when abused, almost all dogs will be horrible pets unless someone puts time and energy into rehabilitating them.

I would agree with that statement sometimes, however it is necessary to look at the specifics. For example, without careful consideration, that line of reasoning will lead to implementation of solutions that seem intuitively helpful (if not ideal) on the surface, but which in reality would be not worth the cost (in dollars or freedom), or might even be outright useless or counter-productive. If you don't mind a few extreme examples to illustrate my point, the fact that people drive too fast in residential areas and thus sometimes kill small children is a problem -- should we ban cars in residential areas? Or what about mandate some sort of speed regulator paired with a built-in GPS device so prevent cars from going about 25mph when driving on residential roads? They are solutions, if not ideal, and so should we not implement one of them rather than do nothing?

It all depends on what you reasonably expect you will accomplish (including foreseeable side effects), and whether the positives are worth the negatives. In the two examples I just gave, I think most people would agree that the negatives outweigh the positives. As someone who generally thinks bureaucracy and legislation have inherent negatives, by default I will reject the idea of banning anything unless I'm given reason to believe it will do some substantial good. Since we're not talking about a specific implementation of how to ban certain breeds, I can't effectively evaluate how much good I think it will do and how much harm -- but since the idea at its core is flawed, I will default to saying it will do more harm than good.

Sure, but do you think the people who have vicious bit-bulls now will trade them for cocker-spaniels if you ban pit-bulls? They will opt for some other medium-large, strong, intelligent, protective breed that just doesn't have the same reputation (yet), against which you would fare no better than the doberman. Or are you suggesting we ban all dogs above 30 lbs?

Well, characterize the problem for me first (I'm not aware of a specific description of the problem that we're discussing at the moment, just the general idea that sometimes dogs attack people). We're not going to pre-emptively prevent all dog attacks, just like we're not going to pre-emptively prevent all car accidents or all violent crime. It's worth doing what we reasonably can, but realistically it's not a problem we can "solve", just something we should try to manage reasonably.

I agree, and maybe should have moderated my statements a bit more to make that clear. You'll note that I did mention that certain personality traits are characteristic of certain breeds, so clearly I agree that temperament is not ENTIRELY a product of their environment. What I intended to convey is that "viciousness" is something which will not occur naturally in a dog without environmental factors playing a significant role, except possibly in extremely rare cases (the equivalent of a human with genetic sociopathy, assuming such a thing exists). If the dog has not been trained well enough for the owner to control it properly, then it should be kept restrained more than most dogs might require (keep it in a large fenced yard instead of taking it to the dog park and letting it run around off the leash), but that doesn't make it a vicious dog, or even dangerous if managed properly.

I can't believe no one has effectively articulated the following yet...

GP said:

Note the quotes around "vicious"? That's because breeds are not inherently vicious, individuals are. Banning certain breeds is neither necessary to eliminate all dangerous dogs (because within any dog breed, there will be well-behaved, friendly dogs), nor is it sufficient (because within any dog breed, there will be mistreated, dangerous dogs). If you speak to a knowledgeable dog owner or breeder, they will tell you that upbringing and environment matters far more than genetics in determining whether a dog will be dangerous/vicious. If you ban certain breeds, you'll be needlessly banning many friendly dogs, and you won't be addressing dangerous dogs of other breeds. If you want to legislate your freedom against dangerous dogs, use a sensible definition of "dangerous", rather than an arbitrary one such as "dogs of the following breeds: ..." If a dog attacks someone without reasonable provocation (as determined by a criminal court), have the dog euthanized, fine the owner, even jail the owner if it can be shown that the owner was responsible due to deliberate actions or gross negligence. In severe cases where an owner is a repeat offender, maybe even prohibit that person from ever owning or being responsible for a dog again. But ban a breed just because the misinformed public believes they are more likely to be vicious? No.

Just to pre-empt some of the expected rebuttals, yes certain dog breeds are responsible for a greater share of dog attacks than others, even after controlling for the number of such dogs in a given area. But banning those breeds won't solve the problem, or even alleviate it in a meaningful way for more than a year or two (time enough for people to acquire another puppy and then train/abuse it into becoming dangerous). Repeat after me: correlation is not causation. The reason those dogs are involved in more attacks is -- get this -- because people think they are more vicious! If you were the sort of person who wants to have a vicious guard dog (which you won't train properly, or might even abuse in order to encourage viciousness), you're going to want a dog that will be perceived by others as big, strong, and dangerous, because that makes a more effective guard dog. It doesn't matter if the dog has any tendency towards viciousness or not, if the dog is anything less than perfectly trusting of strangers, you can make the dog vicious. And then that breed will have more than its fair share of vicious dogs, not because the breed trends that way inherently, but because people who want a vicious dog trend towards buying those breeds! If you ban that breed, it won't stop them from getting a dog and making it dangerous, they'll just do it with a different breed with appropriate characteristics (physically strong, reasonably or highly intelligent, having a strong pack/family sense, and a tendency to be protective of their pack/family -- all positive characteristics in a properly-trained dog). Of course, if enough people pick the same breeds based on those characteristics, then those breeds will gain a reputation for being vicious and get added to the list of banned breeds, rinse, repeat.

Did you seriously just ask for anecdotal evidence? Fine, I'll bite.

A friend of mine was in a horrible car wreck about a month ago -- she was driving a pickup truck on the Interstate at probably 70+mph (the speed limit is 70mph around here), when the brakes suddenly seized up. She lost control of the vehicle, and long story short (as I'll probably get the details wrong if I try to recount exactly what happened), her left wrist was shattered, her scalp almost totally ripped from her skull, and her neck was broken in 4 places. She is alive, and thank God it appears she will suffer no permanent spinal or brain injury. She was *incredibly* lucky that she wasn't killed outright (based on the condition of the passenger side, anyone else in the car wouldn't have lived), much less that she will be able to more-or-less recover with physical therapy over the course of the next few years.

Although much less dramatic, about 8 years ago I was driving my (fairly old) car on the Interstate at probably around 65mph-70mph when the engine died without warning during a blizzard. Luckily, I wasn't in heavy traffic and I tend to react quickly in crises, so I put the car in neutral and was able to pull over to the shoulder without causing a wreck. Being in the middle of a blizzard, the shoulder was so covered in snow that I had no traction once there, and so I did bump against the guardrail rather harder than I would have liked -- cracked some plastic and scrapped off some paint. If there had been a wreck, I wouldn't have considered it "my incompetence", even if I hadn't thought to put the car in neutral. Never did figure out why that happened, the only other time it ever did anything similar was one time the engine cut out after I had pulled over to the shoulder due to such heavy rain that I didn't feel safe driving at any speed -- I guess I was probably lucky it didn't cut out while I was still in the road.

Hmm, that's very interesting. While I am fairly good at "juggling symbols", I am an extremely visual learner. I am talented at math (and double-majored in it as an undergrad), but growing up I would rarely solve word problems in the "expected" way -- I constantly had to ask my math teachers if what I did was also correct. While I don't think my way of doing math is quite as visual as you describe, I do sometimes imagine visual interactions between parts of equations as I manipulate them.

I have always been interested in the ways in which learning takes place (and teaching of course, though I prefer to think of it as facilitation of learning), and have a vague motivation to someday pursue a PhD in something along the lines of Computer Science Education -- research into how best to facilitate learning about computer science. Of course it's not easy to find places which offer such a program, and I don't particularly want to move for such an opportunity.

With all that said, do you have any interest in having someone like me proofread/copyedit your book, when the time comes?

If you look around the Internet though, there are a number of novel symptoms associated with swine flu that nerds might want to know about.

Fight Club?

Exactly. Just yesterday, I was considering taking up the hobby; it's something I've thought about in the past, but never gotten around to doing. If it's looking like I won't even be able to try it out and talk to other ham operators unless/until there is a major emergency, that's certainly not going to encourage me to spend a significant amount of time, space, and money getting into the hobby (if you could even call it that under such circumstances).

Let me start over. First, I see the original task as an attempt to identify clusters (if they exist, which it turns out they do), as those are of most use. I see no reason for the task to be framed as subdividing the player style space into a fully covering set. Would you care to explain that perspective, rather than just saying it's what you think most people would do? The fact is, humans have a tendency to want fully-categorized data; we're not very comfortable with cases that "don't fit", and so we will often engineer that possibility out of existence. But that doesn't make full categorization any less arbitrary, and it doesn't make it the best choice, even if most people would indeed prefer it that way. Interestingly, having now read the research paper, these designers did not force the clusters to cover all players, so they seem to agree with me -- 6.46% of the players were not assigned a cluster, compared to 8.68% in the smallest assigned cluster; this is not an insignificant number.

They are as different as apples and oranges; we seem to be confusing 2 different issues. One is whether it is appropriate to ignore outliers in training of the NN (or defining clusters by some other data clustering algorithm, as you seem to suggest), and the other is what to do with outliers when they are encountered in the game. I see no specific reason to retrain without the outliers, but that's not the issue I was attempting to debate. Regardless of whether you retrain without the outliers, you will still need to decide what to do with the outliers when they are encountered playing the game. Group them into the nearest category (in hopes that the chosen tweaks actually improve gameplay), or use a default gameplay mode, or perhaps something else? Assuming the modes for the identified classes are all tweaked variations on a default mode (which may or may not be true), I see no reason never to use the default mode simply because we like to have fully categorized data. It can maybe be summed up this way: "We'll make the game as we normally do, but identify certain things that can be made easier or more challenging if we have reason to think it will improve the player's experience. If we can't tell confidently whether a given tweak is likely to help or hurt, we'll leave it alone." I think that is a very straightforward, sensible approach; if you disagree, I'm interested in hearing why.

Really? I can accept that you would do that, but it really doesn't seem clear to me that forcing it to pick one is what "most designers" would do, or that it is the best option even if "most designers" would do it.

How would you ignore outliers in your NN? The NN makes the categorizations after looking at all the data, so how can you know what the outliers are until the NN has already considered and incorporated them? You also probably can't arbitrarily "add a new class" without making changes to your initial NN setup and starting the learning process over. You left out an option, the one I think is most likely: apply a default non-class when a categorization can't be made with high confidence.

The way the summary reads, it sounds to me like there is a default gameplay which is then tweaked (probably manually by the game designers) for each group. One of the great things about NN is that they can, if designed properly, give you information about the confidence level of the result rather than just the result. I would presume that it is best to leave the game on the default gameplay setting, and only apply a group-specific tweak if the player clusters with a group fairly strongly, since the summary indicates this would cover the vast majority of people. This is based on the logic that the default gameplay will have the most general appeal, and it is better to go with the default than make a guess with low confidence -- choosing the wrong tweak would likely do more harm to the player's enjoyment than choosing not to tweak at all.