Coming Soon!

[Joy Christian]

Good news! I am able to duplicate the results of all that Mathematica code that does the analysis of the observations (outA and outB) to arrive at outA5 and outB5 with two statements and two simple do loops:

outA5 = outA;
outB5 = outB;

Do[If[outB[[i]][[4]] == f2 && outA[[i]][[2]] != outA[[i]][[5]],
outA5[[i]][[2]] = outA[[i]][[2]]*-1], {i, m}]

Do[If[outA[[i]][[4]] == f1 && outB[[i]][[2]] != outB[[i]][[5]],
outB5[[i]][[2]] = outB[[i]][[2]]*-1], {i, m}]

I am able to duplicate the results of finding CHSH and also the plot of the cosine function.

I was able to do this by taking Fred's code and seeing what it is actually doing. This took me some time, but the final result really quite simple. Now the program is much faster, and by looking at the do loop you can clearly see that the calculation is non-local. The program modifies A depending on values in B and vice versa.

Except that is not my model. It's your model which is not the same as mine. You are not using the trial numbers which are part of the hidden variables since shared by Alice and Bob.

I was about to say the same! Nevertheless, we can use John's modified program if there is anything good in it, if that helps to speed up your (i.e., Fred's) program.
.

[jreed]

Good news! I am able to duplicate the results of all that Mathematica code that does the analysis of the observations (outA and outB) to arrive at outA5 and outB5 with two statements and two simple do loops:

outA5 = outA;
outB5 = outB;

Do[If[outB[[i]][[4]] == f2 && outA[[i]][[2]] != outA[[i]][[5]],
outA5[[i]][[2]] = outA[[i]][[2]]*-1], {i, m}]

Do[If[outA[[i]][[4]] == f1 && outB[[i]][[2]] != outB[[i]][[5]],
outB5[[i]][[2]] = outB[[i]][[2]]*-1], {i, m}]

I am able to duplicate the results of finding CHSH and also the plot of the cosine function.

I was able to do this by taking Fred's code and seeing what it is actually doing. This took me some time, but the final result really quite simple. Now the program is much faster, and by looking at the do loop you can clearly see that the calculation is non-local. The program modifies A depending on values in B and vice versa.

Except that is not my model. It's your model which is not the same as mine. You are not using the trial numbers which are part of the hidden variables since shared by Alice and Bob.
.

Everything in this model is the same as yours, except I replaced all that Mathematica spaghetti code you had written with these simple statements. The output results are identical, and it runs much faster. I can send you the full notebook if you are interested.

[gill1109]

Good news! I am able to duplicate the results of all that Mathematica code that does the analysis of the observations (outA and outB) to arrive at outA5 and outB5 with two statements and two simple do loops:

outA5 = outA;
outB5 = outB;

Do[If[outB[[i]][[4]] == f2 && outA[[i]][[2]] != outA[[i]][[5]],
outA5[[i]][[2]] = outA[[i]][[2]]*-1], {i, m}]

Do[If[outA[[i]][[4]] == f1 && outB[[i]][[2]] != outB[[i]][[5]],
outB5[[i]][[2]] = outB[[i]][[2]]*-1], {i, m}]

I am able to duplicate the results of finding CHSH and also the plot of the cosine function.

I was able to do this by taking Fred's code and seeing what it is actually doing. This took me some time, but the final result really quite simple. Now the program is much faster, and by looking at the do loop you can clearly see that the calculation is non-local. The program modifies A depending on values in B and vice versa.

Congratulations, John! Please send me the notebook and I'll see if I can rewrite it in R.
Except that is not my model. It's your model which is not the same as mine. You are not using the trial numbers which are part of the hidden variables since shared by Alice and Bob.
.

[FrediFizzx]

Everything in this model is the same as yours, except I replaced all that Mathematica spaghetti code you had written with these simple statements. The output results are identical, and it runs much faster. I can send you the full notebook if you are interested.

Nope, of course it is non-local if you don't use the trial numbers for matching. I can do a even a more simple non-local model that is superfast. Here is 20 million trials in 9 minutes with 1 degree resolution. So, don't need to see your code. But you could post it up on the Wolfram Cloud for display and post a link to it.



This was the inspiration for the current local model.

EPRsims/newCS-1-noLoops-non.pdf
EPRsims/newCS-1-noLoops-non.nb
.

[Joy Christian]

.
Richard D. Gill and John Reed are up to their usual trick of producing strawman arguments --- or, in this case, strawman code.

Gill has done that time and again with my work during the past ten years. He has zero credibility in my books.
.

[gill1109]

.
Richard D. Gill and John Reed are up to their usual trick of producing strawman arguments --- or, in this case, strawman code.
Gill has done that time and again with my work during the past ten years. He has zero credibility in my books.

Who cares about my credibility in your books!
Seriously, when you try to get the paper published, you'll be glad that someone has figured out what Fred's code actually does, so you can write it up yourself in an intelligible way.

Obviously the simulation is non-local - it violates Bell's inequality!

[Joy Christian]

.
Richard D. Gill and John Reed are up to their usual trick of producing strawman arguments --- or, in this case, strawman code.
Gill has done that time and again with my work during the past ten years. He has zero credibility in my books.

Who cares about my credibility in your books!
Seriously, when you try to get the paper published, you'll be glad that someone has figured out what Fred's code actually does, so you can write it up yourself in an intelligible way.

Obviously the simulation is non-local - it violates Bell's inequality!

John Reed's simulation is non-local. Why should anyone care about his non-local simulation?
.

[FrediFizzx]

Good news! I am able to duplicate the results of all that Mathematica code that does the analysis of the observations (outA and outB) to arrive at outA5 and outB5 with two statements and two simple do loops:

outA5 = outA;
outB5 = outB;

Do[If[outB[[i]][[4]] == f2 && outA[[i]][[2]] != outA[[i]][[5]],
outA5[[i]][[2]] = outA[[i]][[2]]*-1], {i, m}]

Do[If[outA[[i]][[4]] == f1 && outB[[i]][[2]] != outB[[i]][[5]],
outB5[[i]][[2]] = outB[[i]][[2]]*-1], {i, m}]

I am able to duplicate the results of finding CHSH and also the plot of the cosine function.

I was able to do this by taking Fred's code and seeing what it is actually doing. This took me some time, but the final result really quite simple. Now the program is much faster, and by looking at the do loop you can clearly see that the calculation is non-local. The program modifies A depending on values in B and vice versa.

Except that is not my model. It's your model which is not the same as mine. You are not using the trial numbers which are part of the hidden variables since shared by Alice and Bob.

I was about to say the same! Nevertheless, we can use John's modified program if there is anything good in it, if that helps to speed up your (i.e., Fred's) program.
.

Nope, nothing useful in a non-local model. I already posted a better non-local model that was the inspiration for the local model. Jeez, we gave everyone the analytical formulas for the local model and Gill still doesn't understand it or probably doesn't want to understand it because it kills his fantasy Bell world.
.

[Heinera]

Good news! I am able to duplicate the results of all that Mathematica code that does the analysis of the observations (outA and outB) to arrive at outA5 and outB5 with two statements and two simple do loops:
[...]

Excellent work, John! This is exactly what they are doing. Non-local as bloody hell.

[Joy Christian]

Good news! I am able to duplicate the results of all that Mathematica code that does the analysis of the observations (outA and outB) to arrive at outA5 and outB5 with two statements and two simple do loops:
[...]

Excellent work, John! This is exactly what they are doing. Non-local as bloody hell.

Says Mr. Nobody as bloody hell.
.

[FrediFizzx]

Good news! I am able to duplicate the results of all that Mathematica code that does the analysis of the observations (outA and outB) to arrive at outA5 and outB5 with two statements and two simple do loops:
[...]

Excellent work, John! This is exactly what they are doing. Non-local as bloody hell.

Says Mr. Nobody as bloody hell.
.

Amazing that they don't understand simple analytical formulas.
.

[FrediFizzx]

Good news! I am able to duplicate the results of all that Mathematica code that does the analysis of the observations (outA and outB) to arrive at outA5 and outB5 with two statements and two simple do loops:
[...]

Excellent work, John! This is exactly what they are doing. Non-local as bloody hell.

Says Mr. Nobody as bloody hell.
.

Amazing that they don't understand simple analytical formulas.
.

Oh, better post the link to them as they probably can't even find that.

http://dx.doi.org/10.13140/RG.2.2.28311.91047
.

[gill1109]

Says Mr. Nobody as bloody hell.
.

Amazing that they don't understand simple analytical formulas.
.

Oh, better post the link to them as they probably can't even find that.

http://dx.doi.org/10.13140/RG.2.2.28311.91047
.

There is no formula for k_A nor for k_B in the paper. There is no way to convert the paper's text and formulas into computer code. The simple analytic formulas (ignoring the comments about the undefined k_A and k_B) actually produce correlations which do not violate the Bell-CHSH inequality. The computer code is hardly annotated. There are no variables called something like k_A and k_B in the code.
Please add analytic formulas for k_A and k_B to the paper.

[FrediFizzx]

@gill1109 CHSH = 2.78608 Trial number = k = k_A = k_B. Too complex for you I guess. The trial number in the code is the 3rd element in the 5 element rows of the tables. Look at outA4 here. See the "i" as the 3rd element. That records the trial numbers from 1 to what ever m is set to. It is the index number of the list which is the same as the trial number.
.

[gill1109]

@gill1109 CHSH = 2.78608 Trial number = k = k_A = k_B. Too complex for you I guess. The trial number in the code is the 3rd element in the 5 element rows of the tables. Look at outA4 here. See the "i" as the 3rd element. That records the trial numbers from 1 to what ever m is set to. It is the index number of the list which is the same as the trial number.
.

No, too simplistic for me. First of all, you have an index k running from 1 up to "m", the number of trials, and for each k there are two inputs (settings) and one or two hidden variables (shared between Alice and Bob). On Alice and Bob's sides one can compute various sines and cosines using the local data and the shared data (the hidden variables). You take the results computed on Alice's side, put it in a matrix, and reorder the rows. Notice the Mathematica function "Sort".

You do something similar for the data from Bob's side.

Then you bring the data together and compute the usual statistics. This means that at the last stage, the k'th row of the combined data includes stuff from two different trials - one of Alice's trials, and one of Bob's trials. You could define k_A(k) to be the original trial number of Alice's part of row k of the output data, and k_B(k) to be the original trial number of Bob's part of row k of the output data. Thus k_A and k_B are vectors of "m" elements defining the permutations of Alice and Bob's original data which your matching procedure does.

Define these permutations mathematically. That is the missing part of the present paper. Joy Christian can write the formulas for you once he too has figured out what you are actually doing. John Reed's code does exactly the same thing but is more transparent. Joy might find John's code useful. In the meantime, I will translate John' Mathematica code to R so I too can better understand what you are doing. Then we can compare your results to recent theoretical results concerning how much non-local communication is necessary to convert local and realistic data into data that imitates quantum predictions.

I think I will also be able to do an analytical calculation of what your simulation is doing and improve it so as to *exactly* reproduce the negative cosine in the limit of large "m"

[FrediFizzx]

@gill1109 Every thing I do is already defined in the paper. Reed's code is not the same as what I do. If you follow that, it will be a strawman. A non-local one. It would be easier to just do the non-local sim I did for a non-local simulation in R. Simple "or" functions. There is no particular "permutations". If you have a question about the code, just ask it. Most of the code is pretty self explanatory. So, if you tell me what you don't understand then I will annotate that part.
.

[Justo]

@gill1109 Every thing I do is already defined in the paper. Reed's code is not the same as what I do. If you follow that, it will be a strawman. A non-local one. It would be easier to just do the non-local sim I did for a non-local simulation in R. Simple "or" functions. There is no particular "permutations". If you have a question about the code, just ask it. Most of the code is pretty self explanatory. So, if you tell me what you don't understand then I will annotate that part.
.

I am not trying to follow the code. I just want to understand the model from what is explained in the introductory description. There is one thing that is left unexplained.
The experiment requires only particles that originate in the same event (k_A=k_B) be considered. Why would you consider cases where , it seems to me that case should not happen in a theoretical simulation. In any case, since those cases are included, it should be explained why. It seems that there is a hidden phenomenon that forces us to consider cases with different numbers of trials.
Whatever it is that produces coincidences or not coincidences is a hidden variable too. It is important to know whether that is a local phenomenon.
Of course, if one analyzes the code, one could figure out what is going on. However, if the paper is intended for a general audience and, especially if it is intended for publication, that point should be explained because it is not evident.

[FrediFizzx]

@gill1109 Every thing I do is already defined in the paper. Reed's code is not the same as what I do. If you follow that, it will be a strawman. A non-local one. It would be easier to just do the non-local sim I did for a non-local simulation in R. Simple "or" functions. There is no particular "permutations". If you have a question about the code, just ask it. Most of the code is pretty self explanatory. So, if you tell me what you don't understand then I will annotate that part.
.

I am not trying to follow the code. I just want to understand the model from what is explained in the introductory description. There is one thing that is left unexplained.
The experiment requires only particles that originate in the same event (k_A=k_B) be considered. Why would you consider cases where , it seems to me that case should not happen in a theoretical simulation. In any case, since those cases are included, it should be explained why. It seems that there is a hidden phenomenon that forces us to consider cases with different numbers of trials.
Whatever it is that produces coincidences or not coincidences is a hidden variable too. It is important to know whether that is a local phenomenon.
Of course, if one analyzes the code, one could figure out what is going on. However, if the paper is intended for a general audience and, especially if it is intended for publication, that point should be explained because it is not evident.

Hmm... I think I already explained it. Perhaps you missed it. The explanation is when you split the events into A_1 and A_2 which are events greater than HV and less than HV respectively (BTW, we are adding to those two lines If functions, "otherwise no result" so it is better defined). You find that there are events in A_1 that don't have a matching event in B_1 and in B_1 there are events where no match in A_1. So, those events are taken out of A_1 and B_1 then the result is the rest of the events in A_1 match those in B_1. You might think that process is non-local but its not since we only use the shared trial numbers between A and B to accomplish it. Now, what do you think should happen to those events that were pulled out?
.

[gill1109]

@gill1109 Every thing I do is already defined in the paper. Reed's code is not the same as what I do. If you follow that, it will be a strawman. A non-local one. It would be easier to just do the non-local sim I did for a non-local simulation in R. Simple "or" functions. There is no particular "permutations". If you have a question about the code, just ask it. Most of the code is pretty self explanatory. So, if you tell me what you don't understand then I will annotate that part.

John Reed verified that his code produced identical results to yours. I have further simplified his code and rewritten it in R.

Code: Select all

M <- 100000
set.seed(1234)
aBits <- sample(c(0, 1), size = M, replace = TRUE)
bBits <- sample(c(0, 1), size = M, replace = TRUE)
aRadians <- aBits * pi / 2
bRadians <- bBits * pi / 2 + pi / 4
set.seed(5678)
R <- 0.25 * cos(runif(M, 0, pi))^2  # Fodje's M hidden variables rho
S <- runif(M, 0, 2*pi)  # Angles of M spin vectors (radians)
C <- abs(cos(aRadians - S)) < R        # TRUE is "fail", FALSE is "good"
A <- ifelse(C, -sign(sin(aRadians - S)), -sign(cos(aRadians - S)))
AA <- -sign(sin(aRadians - S))
D <- abs(cos(bRadians - S)) < R        # TRUE is "fail", FALSE is "good"
B <- ifelse(D, +sign(sin(bRadians - S)), +sign(cos(bRadians - S)))
BB <- +sign(sin(bRadians - S))
A <- ifelse(D & !(A == AA), -A, A)
B <- ifelse(C & !(B == BB), -B, B)
mean((A*B)[aBits == 0 & bBits == 0])
mean((A*B)[aBits == 0 & bBits == 1])
mean((A*B)[aBits == 1 & bBits == 0])
mean((A*B)[aBits == 1 & bBits == 1])
- mean((A*B)[aBits == 0 & bBits == 0]) + mean((A*B)[aBits == 0 & bBits == 1]) -
          mean((A*B)[aBits == 1 & bBits == 0]) - mean((A*B)[aBits == 1 & bBits == 1])

The result was S = 2.63625

I noticed that Fred's code used the function "Sort" which one would expect would produce a permutation, but John Reeds didn't. You can see the nonlocality in the lines where the sign of the outcomes A and B are switched depending in part on information from the other wing of the experiment - Fred's C1 and C2 which I now called C and D.

I should imagine that you would get better results using Pearl's model rather than Michel Fodje's. It should now be easy to compute analytically what is going on, and also to compute how many non-local sign switches are performed.

If anyone disagrees I suggest they do their own conversion of Fred's Mathematica code to R or Python, and while they're at it, remove superfluous objects leading to superfluous copyings and renamings which only make the code harder to read.

Note: the R function ifelse(test, x, y) evaluates "test" and returns "x" or "y" depending on whether "test" delivers TRUE or FALSE. It works on vectors elementwise, delivering a vector of results of the same length as test, x and y. Notice that I originally chose binary settings completely at random, later converting them to the standard angles measured in radians. I chose the spin vector uniformly at random on the circle, rather than choosing it at random as a whole number of degrees.

[Joy Christian]