The real puzzle

Foundations of physics and/or philosophy of physics, and in particular, posts on unresolved or controversial issues

Re: The real puzzle

Postby Heinera » Sun May 04, 2014 7:49 am

minkwe wrote:Unfortunately, you are just wrong. Statistics is irrelevant. I've challenged you in the past to produce an appropriate data set from any source, which violates the appropriate upper bound by even 0.0000000000001 statistically. You have been unable. Keyword is appropriate. Introduce as much error as you like.

Huh? Didn't the following post violate the upper bound of 2?
gill1109 wrote:And here are some new results, with many thanks to Zen for new ideas and new result: running [Joy's] program *four* times, to make four separate files, each analysed separately, the four correlations were:

-0.75156, 0.24978, -0.75114, -0.24862

whose absolute values add up to 2.0011. Not exactly 2. But not far off (about 1 / sqrt N off target?).

Or do you mean that in this case the appropriate upper bound is 4?

Then you would be correct if you with upper bound meant "absolute, logically possible upper bound".

But the correct meaning of upper bound in this case is "upper bound for expectations", which implies that for a LHV model, any given value strictly larger than 2 becomes more and more unlikely as the number of runs N goes toward infinity. You say that statistics is irrelevant, but only by applying statistics will it be possible to distinguish between QM predictions and a LHV model for an experiment with a finite number of runs.
Heinera
 
Posts: 917
Joined: Thu Feb 06, 2014 1:50 am

Re: The real puzzle

Postby minkwe » Sun May 04, 2014 8:09 am

gill1109 wrote:
minkwe wrote:Interestingly, your LG paper, while attempting to rescue Bell, actually confirms its death. One may ask, what does QM say about delta or gamma? Adeniers paper answers that. There are only two possibilities. Either gamma is 1 (strongly objective) or gamma is zero ( weakly objective).

The whole point is that gamma is observable. The right hand side of our inequality is just as experimentally accessible as the left hand side. In real experiments, gamma might be anything between 0 and 1.
.

Please read the statement you are responding to. It's asking about QM not experiment. What in your opinion does QM say about gamma. Note, your answer, combined with your LG result will tell you the *appropriate* upper bound to use. And you will find that QM does not violate it.
Last edited by minkwe on Sun May 04, 2014 8:16 am, edited 1 time in total.
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby minkwe » Sun May 04, 2014 8:14 am

Heinera wrote:Huh? Didn't the following post violate the upper bound of 2?

Heinera, you do not understand anything I'm talking about. What was the degree of disjointedness in your quoted experiment. That will give you the APPROPRIATE upper bound. Try to understand before you "huh?". This is not my result. This is Richards theorem from his LG paper.
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby Heinera » Sun May 04, 2014 8:27 am

minkwe wrote:
Heinera wrote:Huh? Didn't the following post violate the upper bound of 2?

Heinera, you do not understand anything I'm talking about.

In fact, I'm beginning to think I do. And I think our differences can be summed up in this little imaginary exchange:

We have a coin together. I want to find out if the coin is fair or not. We throw it 1000 times, and it comes up with 800 heads and 200 tails.

Me: "S***. This coin is biased."
You: "How do you know? 800 heads is a logically possible outcome! In fact, the only upper bound is 1000 heads."
Me: "Trust me, the coin is biased."

(I even managed to give myself the last word here... :D )
Heinera
 
Posts: 917
Joined: Thu Feb 06, 2014 1:50 am

Re: The real puzzle

Postby minkwe » Sun May 04, 2014 8:30 am

Heinera wrote:But the correct meaning of upper bound in this case is "upper bound for expectations", which implies that for a LHV model, any given value strictly larger than 2 becomes more and more unlikely as the number of runs N goes toward infinity.

Wrong! Look at Richards result. The appropriate upper bound " for expectations " can be anything from 2 to 4, depending on degree of disjointedness. Whether N is 5 or 10 million.
You say that statistics is irrelevant, but only by applying statistics will it be possible to distinguish between QM predictions and a LHV model for an experiment with a finite number of runs.

Wrong again. Only by comparing QM and its appropriate upper bound, and comparing the LHV model with its appropriate upper bound will you be making any sense at all. Anything else is nonsense.
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby minkwe » Sun May 04, 2014 8:48 am

Heinera wrote:In fact, I'm beginning to think I do. And I think our differences can be summed up in this little imaginary exchange:

We have a coin together. I want to find out if the coin is fair or not. We throw it 1000 times, and it comes up with 800 heads and 200 tails.

Me: "S***. This coin is biased."
You: "How do you know? 800 heads is a logically possible outcome! In fact, the only upper bound is 1000 heads."
Me: "Trust me, the coin is biased."

(I even managed to give myself the last word here... :D )


Nope. I already gave you a coin toss example which captures the issue here :
viewtopic.php?f=6&t=44

Yours is a joke. You still didn't say what the degree of "disjointedness" was in your quoted example.

The LG theorem is proving to be useful after all as a way if determining the appropriate upper bound. You will find that the appropriate upper bound is never violated, even for QM.
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby minkwe » Sun May 04, 2014 9:05 am

My challenge for Richard and his friends is still open:

I challenge you to produce the non-real/non-local dataset from any source, including as much error as you like, which violates the appropriate bound by even 0.000001. You can even use QM if you can. No restrictions on N.

We will use Richards LG theorem to calculate the appropriate upper bound, based on the degree if disjointedness in your dataset.
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby gill1109 » Sun May 04, 2014 9:52 am

minkwe wrote:
gill1109 wrote:
minkwe wrote:Interestingly, your LG paper, while attempting to rescue Bell, actually confirms its death. One may ask, what does QM say about delta or gamma? Adeniers paper answers that. There are only two possibilities. Either gamma is 1 (strongly objective) or gamma is zero ( weakly objective).

The whole point is that gamma is observable. The right hand side of our inequality is just as experimentally accessible as the left hand side. In real experiments, gamma might be anything between 0 and 1.
.

Please read the statement you are responding to. It's asking about QM not experiment. What in your opinion does QM say about gamma. Note, your answer, combined with your LG result will tell you the *appropriate* upper bound to use. And you will find that QM does not violate it.

What Adenier apparently said is totally irrelevant.

And the question as to what QM predicts here, is (almost) totally irrelevant too.

LG derive a relation which, in certain well prescribed circumstances, and in the large N limit, will hold between four observable correlations and the minimum of four observable probabilities, if the physics behind the experiment can be described by LHV. We also make perfectly clear what we mean by that.

Why does QM have to say anything about gamma? Well ... and this is where Bell's fifth position comes in ... it might just be the case that if we really did all the QM carefully not just of the two measurement outcomes but also of the two measurement times ... remember the uncertainty principle? ... we would find that the same relation holds, but now assuming QM instead of LHV.

But no-one has done the tricky QM (though Khrennikov and Volovic did do something going a little bit in the right direction) so no-one knows. Emilio Santos has a whole list of arguments too. I understand however that the mainstream QM experts think this is a totally absurd suggestion. They are pretty sure that QM can beat the bound here, as long as gamma is so large, that the bound (the number on the right hand side) is good bit smaller than 2 sqrt 2. (But maybe the mainstream opinion is wrong! It has happened before! It happens - in dramatic form - perhaps once a century.)

Anyway, you don't have to do the tricky QM in order to actually do the experiment with allegedly entangled photons etc and see if you can get to see four correlations and four detection probabilities such that the LHV bound is resoundingly violated - ie with a whole heap of standard errors difference! The experimenters say that they are nearly there. I'm waiting. They know exactly what they have to do. They think it is "just" a question of refining all the stuff which they are already using. Better detectors, faster switches, more expensive crystals for the parametric down conversion, much faster electronics, more careful tuning, and so on and so forth... in other words "merely" quantitative improvement. No essentially new technology. No quantum jump in concepts and techniques.
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden

Re: The real puzzle

Postby gill1109 » Sun May 04, 2014 10:07 am

minkwe wrote:My challenge for Richard and his friends is still open:

I challenge you to produce the non-real/non-local dataset from any source, including as much error as you like, which violates the appropriate bound by even 0.000001. You can even use QM if you can. No restrictions on N.

We will use Richards LG theorem to calculate the appropriate upper bound, based on the degree if disjointedness in your dataset.

You had better tell me and my friends what the appropriate upper bound is. Didn't you say it was 4? Or do you mean, that I should give a data-set which violates the bound written down in LG, but with theoretical population means (expectation values) and theoretical population probabilities, replaced by sample averages and sample relative frequencies respectively? Or do you mean that the appropriate upper bound, by definition, is the smallest bound which can never ever be violated? So you are challenging me to do something which by definition is impossible? I thought there were not going to be any more silly games.

Whichever of those three choices, the challenge is either impossible or trivially easy to satisfy, so not worth doing. However if you want to challenge me to do something which is possible and if you're offering hard cash for the winner, then I might be interested.

(This reminds me of another challenge which is presently taxing the mind of our friend JJC from Oxford, UK. Especially the variant: challenge to do something which is impossible by definition).

I already said that I had no idea what the real puzzle is to which this topic is devoted, but I do understand that you are very puzzled. There is no place for statistics in your scientific universe, no place for thought experiments, no place for mathematical theorems (since they are tautologies, hence superfluous). There is no time to read Bell, read a text book on applied statistics, or try out a pedagogical R experiment. What is there left? Just past experiments, and computer scripts which simulate the data of past experiments. How will science progress if that is all there is? It is very puzzling indeed.

What is the point of talking to someone who already knows everything that they will ever, ever need to know? Who never makes any attempt to see things from someone else's point of view, and always insists that his point of view is the only valid point of view?
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden

Re: The real puzzle

Postby minkwe » Sun May 04, 2014 11:33 am

gill1109 wrote:You had better tell me and my friends what the appropriate upper bound is. Didn't you say it was 4?

Haven't you learnt already that those silly games may work on others but not me?

You claim that QM can violate a bound. Or that non-local theories can violate a bound. Or that non-real theories can violate a bound. And that LHV can not violate said bound. You further claim that a the bound can be violated statistically by experimental error. I ask you to provide the QM/non-local/non-real/error-prone dataset which violates the bound.
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby minkwe » Sun May 04, 2014 11:38 am

gill1109 wrote:Or do you mean, that I should give a data-set which violates the bound written down in LG

LG does not give a bound. It gives a method of calculating an appropriate bound for the given degree of disjointedness. This is your paper not mine. All you have to do is figure out delta, and you have the appropriate bound for any situation. Isn't that your theorem, or do you now want to disown it too. If delta is zero, it means you have a completely disjoint dataset and the absolute upper bound is 4. If delta is 1, you have a single set and the upper bound is 2, just as I've been arguing all along. So the challenge is to produce your dataset from any source whatsoever, including as much error as you like, with any delta you like, which violates the appropriate bound calculated according to your theorem. Can you do that, or are you now ready to admit that the impossibility is universal and applies just as well to local theories as non-local and non-real ones (whatever that means). Not just to LHV as you go around claiming.

Again, in case you don't figure out what the point of the challenge is, let me remind you: You claim that the impossibility of violating the bound is just a feature of LHV models, and I claim that the impossibility is mathematical, nothing whatsoever can violate it, not non-locality, not non-realism, and not even QM. Get it? Forget statistics, it is irrelevant. Although you are free to use as much statistics as you like to produce the dataset which violates the appropriate bound (calculated according to your LG theorem), by just 0.00001 (a much more generous margin than the 0.1 your other bets are based on).
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby gill1109 » Sun May 04, 2014 3:27 pm

minkwe wrote:
gill1109 wrote:Or do you mean, that I should give a data-set which violates the bound written down in LG

LG does not give a bound. It gives a method of calculating an appropriate bound for the given degree of disjointedness. This is your paper not mine. All you have to do is figure out delta, and you have the appropriate bound for any situation. Isn't that your theorem, or do you now want to disown it too. If delta is zero, it means you have a completely disjoint dataset and the absolute upper bound is 4. If delta is 1, you have a single set and the upper bound is 2, just as I've been arguing all along. So the challenge is to produce your dataset from any source whatsoever, including as much error as you like, with any delta you like, which violates the appropriate bound calculated according to your theorem. Can you do that, or are you now ready to admit that the impossibility is universal and applies just as well to local theories as non-local and non-real ones (whatever that means). Not just to LHV as you go around claiming.

Again, in case you don't figure out what the point of the challenge is, let me remind you: You claim that the impossibility of violating the bound is just a feature of LHV models, and I claim that the impossibility is mathematical, nothing whatsoever can violate it, not non-locality, not non-realism, and not even QM. Get it? Forget statistics, it is irrelevant. Although you are free to use as much statistics as you like to produce the dataset which violates the appropriate bound (calculated according to your LG theorem), by just 0.00001 (a much more generous margin than the 0.1 your other bets are based on).

Michel, the bound in our paper is a bound on expectation values. It is not about data-sets. I still have no idea what you want me to do.

I can't submit a counterexample to a true theorem.

I can submit data such that, if you calculate in the obvious way estimates of gamma and S from the data, those numbers violate the inequality. But the theorem does not apply to sample averages. It is not about sample averages. It is about population means.

According to QM, population means can violate the inequality (if the upper bound in the inequality is less that 2 sqrt 2). Obviously data could violate the inequality much more still...
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden

Re: The real puzzle

Postby minkwe » Sun May 04, 2014 7:35 pm

minkwe wrote:So the challenge is to produce your dataset from any source whatsoever, including as much error as you like, with any delta you like, which violates the appropriate bound calculated according to your theorem. Can you do that, or are you now ready to admit that the impossibility is universal and applies just as well to local theories as non-local and non-real ones (whatever that means). Not just to LHV as you go around claiming.

Again, in case you don't figure out what the point of the challenge is, let me remind you: You claim that the impossibility of violating the bound is just a feature of LHV models, and I claim that the impossibility is mathematical, nothing whatsoever can violate it, not non-locality, not non-realism, and not even QM. Get it? Forget statistics, it is irrelevant. Although you are free to use as much statistics as you like to produce the dataset which violates the appropriate bound (calculated according to your LG theorem), by just 0.00001 (a much more generous margin than the 0.1 your other bets are based on).

You know exactly what I'm asking. It is clearly stated above.
Last edited by FrediFizzx on Sun May 04, 2014 9:26 pm, edited 2 times in total.
Reason: Repaired quoting
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby gill1109 » Sun May 04, 2014 10:48 pm

minkwe wrote:
minkwe wrote:So the challenge is to produce your dataset from any source whatsoever, including as much error as you like, with any delta you like, which violates the appropriate bound calculated according to your theorem. Can you do that, or are you now ready to admit that the impossibility is universal and applies just as well to local theories as non-local and non-real ones (whatever that means). Not just to LHV as you go around claiming.

Again, in case you don't figure out what the point of the challenge is, let me remind you: You claim that the impossibility of violating the bound is just a feature of LHV models, and I claim that the impossibility is mathematical, nothing whatsoever can violate it, not non-locality, not non-realism, and not even QM. Get it? Forget statistics, it is irrelevant. Although you are free to use as much statistics as you like to produce the dataset which violates the appropriate bound (calculated according to your LG theorem), by just 0.00001 (a much more generous margin than the 0.1 your other bets are based on).

You know exactly what I'm asking. It is clearly stated above.

Please do it for me. Here are your instructions:

Simulate data from the model explicitly described in the last section of LG with N (number of pairs of emitted particles) equal, say, to 10 000 = 4 * 2 500. Calculate the empirical estimates of gamma (the minimum of four sample relative frequencies) and of the four correlations (four sample averages). You'll find that the inequality which, according to the paper, is proven to hold for expectation values and probabilities is "broken" about 50% of the time by sample averages and sample relative frequencies in simulated data-sets.

I suggest you repeat the basic simulation about 1000 times and plot a histogram of the relevant quantity (empirical counterpart of right hand side minus left hand side of inequality). You can go on to repeat the whole thing with several different values of N and observe how the distribution gets tighter as N grows, at one over square root of N rate.

OK: now I have made some predictions. You can now do the experiment in order to prove I'm wrong.
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden

Re: The real puzzle

Postby minkwe » Mon May 05, 2014 7:43 am

gill1109 wrote:Please do it for me. Here are your instructions:

Won't work. It is my challenge to you, not the other way around.
1) You claim that your LG paper prescribes the appropriate bound that should be used based on the degree of disjointedness "delta". I agree.
2) You claim that an experiment producing a dataset which violates the bound might be possible in the near future. I disagree.
3) You claim that violation of such a bound is a privileged property of QM/non-local theories/non-real theories but impossible for LHV models. I disagree, I claim that nothing whatsoever can violate the bound, unless a mathematical error is being made. I'm challenging you to simulate the non-local/non-real theories which violate the bound as you have been claiming, including in your most recent paper.

I'm simply turning the tables on you. You have various bets asking people to simulate a local theorem which violates some given bound. I'm asking you to simulate a QM/non-local/non-real theorem which violates the bound. This is very clear, so don't pretend not to understand it.

As clear as a bell (pun intended). Provide your dataset, generated using any QM technique, any non-local technique, and any non-real technique, introduce as much experimental/statistical error as you like. Note, unlike you, I have placed no roadblocks of alleged "loopholes" on your way. No restriction on now small or now large N might be. No restrictions about random number seeds. No restrictions on amount of noise or error. AND, no demands about obtaining cosine-squared curves, you can generate any shape of curve you like. AND, you only have to violate the appropriate bound by 0.000001. You are free to simulate it any way you like. And we will use your own theory, to calculate the appropriate bound.

If as you go around claiming, realism is untenable, it means you believe non-realism is tenable. It means you believe the data from experiments and your so-called future successful Bell experiment will be non-real. Now we don't have to wait for the experiment, you can simply write an R-script simulating your non-real dataset which violates the bound. No need to continue bobbing and weaving about statistics and sample/population averages. It's time to put up or shut up. Incorporate all your claims about statistics into your R-script that you produce. You are permitted to any statistics you choose in producing the dataset. You know exactly what I'm asking.
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby gill1109 » Mon May 05, 2014 9:22 am

minkwe wrote:I'm simply turning the tables on you.

Yes that's a fun idea, however the situation is not quite symmetric. I suggest you do my little R experiment first, in order to get me into a more favourable mood.

Your challenge is amusing, I have told you how I would solve it. If you want to get me really interested in it, you had better try out my solution and tell me if it didn't work. If I have understood your challenge correctly, it will work. And seriously, I do think you need to learn some statistics and to read the more mature works of Bell where he carefully filled in the short-cuts in his earlier work.

But I really must get on with some other rather urgent work at the moment. And after that take a quite long vacation as well.
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden

Re: The real puzzle

Postby minkwe » Mon May 05, 2014 10:34 am

gill1109 wrote:
minkwe wrote:I'm simply turning the tables on you.

I suggest you do my little R experiment first, in order to get me into a more favourable mood.


I don't care about your mood. I care about the false claims you go around making in forums and in your papers that QM can do something which LHV theories can not. You haven't provided a solution. Where is the dataset? You keep deluding yourself about statistics, now is the time to put up or shut up.
Here is the challenge again. Read it carefully, and provide the dataset, or by failing to do so, admit that not even QM/non-locality/non-realism can violate the appropriate bound according to your LG theorem.
1) You claim that your LG paper prescribes the appropriate bound that should be used based on the degree of disjointedness "delta". I agree.
2) You claim that an experiment producing a dataset which violates the bound might be possible in the near future. I disagree.
3) You claim that violation of such a bound is a privileged property of QM/non-local theories/non-real theories but impossible for LHV models. I disagree, I claim that nothing whatsoever can violate the bound, unless a mathematical error is being made. I'm challenging you to simulate the non-local/non-real theories which violate the bound as you have been claiming, including in your most recent paper.


What is it going to be, "put up" or "shut up"? Silly games won't work this time, I don't need an explanation from you. I need a non-real/non-local/QM DATASET or any dataset you like with as much statistical error as you like, which violates the appropriate bound by 0.00001. Are you able to do that? Or do you admit, as I've been arguing, that it is impossible.
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby gill1109 » Mon May 05, 2014 5:26 pm

minkwe wrote:What is it going to be, "put up" or "shut up"? Silly games won't work this time, I don't need an explanation from you. I need a non-real/non-local/QM DATASET or any dataset you like with as much statistical error as you like, which violates the appropriate bound by 0.00001. Are you able to do that? Or do you admit, as I've been arguing, that it is impossible.

I already told you how to do it and suggested you find out for yourself that my solution works. You can program, rather well I think.

How much money are you offering?
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden

Re: The real puzzle

Postby minkwe » Mon May 05, 2014 10:19 pm

gill1109 wrote:
minkwe wrote:What is it going to be, "put up" or "shut up"? Silly games won't work this time, I don't need an explanation from you. I need a non-real/non-local/QM DATASET or any dataset you like with as much statistical error as you like, which violates the appropriate bound by 0.00001. Are you able to do that? Or do you admit, as I've been arguing, that it is impossible.

I already told you how to do it and suggested you find out for yourself that my solution works. You can program, rather well I think.

How much money are you offering?

Where is the dataset, Bob N. Weaver?
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: The real puzzle

Postby gill1109 » Mon May 05, 2014 10:35 pm

minkwe wrote:Where is the dataset, Bob N. Weaver?

How much money is in your wallet?
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden

PreviousNext

Return to Sci.Physics.Foundations

Who is online

Users browsing this forum: Majestic-12 [Bot] and 79 guests

cron
CodeCogs - An Open Source Scientific Library