Re: [math-fun] triangular tables [Was: Atrocious Mma ...]
Fred> They've been lazy, that's all: they have assumed that symmetry applies across the board, overlooking the fact that the basic recursion which they have broken at the origin is implied by a global property of the Gamma function; while symmetry is not. rwg>No, it used to work! 9.0.1 Doc: As a bivariate function, Binomial is not continuous in both variables at negative integers: In[1]:= Binomial[-3, -5] Out[1]= 0 In[2]:= Binomial[-3.0000001, -5.0000001] Out[2]= 6. They up and broke it! (And forgot to break the Doc.) Probably in response to a misguided complaint. That's not lazy--that's industrious. Re misguided complaints, I've probably related here Petr Beckmann's unfortunate elision from his History of π, after a reader numerically determined that Product 1/(1-p^-2), p=2,3,5,7,9,..., was not π^2/6. The book never said 9. --rwg Fred> What intrigues me is that their identity reduction mechanism seems to give results correct also for integer arguments, despite the fact that they presumably must invoke properties of Gamma and hypergeometric functions which fail at the singularities. I conjecture that any identity supported in this fashion must hold for all flavours of pseudo-binomial coefficients defined by approaching the limit from a consistent direction. An example would be Rich's one plus two halves-wedge variant, where the limit is approached along x = 2 y rather than y = 0 . Which just goes to show that you can have half your cake and eat it! Fred If true, this should allow such identities to be rewritten generalized. On 7/10/13, Bill Gosper <billgosper@gmail.com> wrote: rwg>Here is an almost correct technique for printing number triangles, of which Julian must repeatedly remind me: pt[n_Integer, from_Integer: 0] := TableForm[Table[If[EvenQ[i + j], "", Binomial[-1 + i, -Floor[n/4] + (1 + i + j)/2] /. 0 -> ""], {i, from, n}, {j, -Ceiling[3*n/4], n + 1}]] [...] Duh, for actual symmetrical triangles, In[516]:= Column[Table[Row[Table[Binomial[n,k],{k,0,n}]," "],{n,0,5}],Center] Out[516]= 1 1 1 1 2 1 1 3 3 1 1 4 6 4 1 1 5 10 10 5 1 --rwg I somehow missed the distressing answer that Maple has been bodysnatched by the same brain virus (or is it Toxoplasma?) that's corrupted Mma: http://isc.carma.newcastle.edu.au/standardCalc accepts Maple input. Standard lookup results for *Pi^binomial(-2,-6)* Best guess: Pi^(5) Both leading CASs! Maybe they think it's just a matter of personal preference, like whether toilet paper spools off the front or the back?
<< As a bivariate function, Binomial is not continuous in both variables at negative integers: In[1]:= Binomial[-3, -5] Out[1]= 0 In[2]:= Binomial[-3.0000001, -5.0000001] Out[2]= 6. They up and broke it! (And forgot to break the Doc.) Probably in response to a misguided complaint. That's not lazy--that's industrious. >> And for their example, they happened to hit upon the precise direction of approach misleadingly but plausibly suggesting symmetry. Textbook victims of Murphy's law, with whom I have to sympathise, on the grounds that similarly misguided faith in the randomness of the universe regularly overtakes me too. WFL On 7/10/13, Bill Gosper <billgosper@gmail.com> wrote:
Fred> They've been lazy, that's all: they have assumed that symmetry applies across the board, overlooking the fact that the basic recursion which they have broken at the origin is implied by a global property of the Gamma function; while symmetry is not.
rwg>No, it used to work! 9.0.1 Doc:
As a bivariate function, Binomial is not continuous in both variables at negative integers:
In[1]:= Binomial[-3, -5]
Out[1]= 0
In[2]:= Binomial[-3.0000001, -5.0000001]
Out[2]= 6.
They up and broke it! (And forgot to break the Doc.) Probably in response to a misguided complaint. That's not lazy--that's industrious.
Re misguided complaints, I've probably related here Petr Beckmann's unfortunate elision from his History of π, after a reader numerically determined that Product 1/(1-p^-2), p=2,3,5,7,9,..., was not π^2/6. The book never said 9. --rwg
Fred> What intrigues me is that their identity reduction mechanism seems to give results correct also for integer arguments, despite the fact that they presumably must invoke properties of Gamma and hypergeometric functions which fail at the singularities. I conjecture that any identity supported in this fashion must hold for all flavours of pseudo-binomial coefficients defined by approaching the limit from a consistent direction.
An example would be Rich's one plus two halves-wedge variant, where the limit is approached along x = 2 y rather than y = 0 . Which just goes to show that you can have half your cake and eat it!
Fred
If true, this should allow such identities to be rewritten generalized.
On 7/10/13, Bill Gosper <billgosper@gmail.com> wrote:
rwg>Here is an almost correct technique for printing number triangles, of which Julian must repeatedly remind me: pt[n_Integer, from_Integer: 0] := TableForm[Table[If[EvenQ[i + j], "", Binomial[-1 + i, -Floor[n/4] + (1 + i + j)/2] /. 0 -> ""], {i, from, n}, {j, -Ceiling[3*n/4], n + 1}]] [...] Duh, for actual symmetrical triangles, In[516]:= Column[Table[Row[Table[Binomial[n,k],{k,0,n}]," "],{n,0,5}],Center] Out[516]= 1 1 1 1 2 1 1 3 3 1 1 4 6 4 1 1 5 10 10 5 1 --rwg I somehow missed the distressing answer that Maple has been bodysnatched by the same brain virus (or is it Toxoplasma?) that's corrupted Mma: http://isc.carma.newcastle.edu.au/standardCalc accepts Maple input. Standard lookup results for *Pi^binomial(-2,-6)* Best guess: Pi^(5) Both leading CASs! Maybe they think it's just a matter of personal preference, like whether toilet paper spools off the front or the back?
_______________________________________________ math-fun mailing list math-fun@mailman.xmission.com http://mailman.xmission.com/cgi-bin/mailman/listinfo/math-fun
participants (2)
-
Bill Gosper -
Fred lunnon