On Sun, Nov 6, 2011 at 4:30 PM, Bill Gosper <billgosper@gmail.com> wrote:
JZucker>
Well, the Calkin-Wilf tree is still the best, of course.
Thank you, Joshua! I had somehow lived in total ignorance of Newman's amazing, useful, and tiny invention: NestList[1/(2*Floor[#] - # + 1) &, 0, 69]
1 1 3 2 1 4 3 5 2 5 3 1 5 4 7 {0, 1, -, 2, -, -, -, 3, -, -, -, -, -, -, -, 4, -, -, -, -, 2 3 2 3 4 3 5 2 5 3 4 5 4 7 3
3 8 5 7 2 7 5 8 3 7 4 1 6 5 9 4 11 -, -, -, -, -, -, -, -, -, -, -, 5, -, -, -, -, --, --, 8 5 7 2 7 5 8 3 7 4 5 6 5 9 4 11 7
7 10 3 11 8 13 5 12 7 9 2 9 7 12 5 --, --, --, --, --, --, --, --, -, -, -, -, --, --, --, 10 3 11 8 13 5 12 7 9 2 9 7 12 5 13
13 8 11 3 10 7 11 4 9 5 1 7 6 11 5 --, --, --, --, --, --, --, -, -, -, 6, -, -, --, --, --, ...} 8 11 3 10 7 11 4 9 5 6 7 6 11 5 14
(http://oeis.org/A002487). Julian derived the reverse sequence: NestList[2*Ceiling[1/#] - 1 - 1/# &, 14/9, 9]
14 5 11 6 7 1 5 9 4 {--, --, --, --, -, -, 6, -, -, -} 9 14 5 11 6 7 6 5 9
and found the waiting time from 355/113 to 0 was 67107848. I.e., intermediate swell was >>355.
Following Josh's pointer http://mathlesstraveled.com/2009/10/18/the-hyperbinary-sequence-and-the-calk... In[1]:= alf[1] = 1; alf[x_ /; x < 1] := 2*alf[x/(1 - x)]; alf[x_] := 1 + 2*alf[x - 1] In[2]:= Timing[alf[355/113]] Out[2]= {0.000129, 67107847} As Yorgey says: Sweet. --rwg
But the real surprise came with plotting the forward sequence starting with Sqrt[-1] in a neighborhood of the origin--the Ford Pinto circles <http:gosper.org/fordpinto.png>! (Complete with some particulate emissions.) (Using Mma's convention: Floor[z]==Floor[Re[z]]+I*Floor[Im[z]].) The reverse sequence plots the left half (mirror image). --rwg rwg>
Duh, the original argument was < -1, so the z/(z-1) transformation gives a single, convergent series:
10 3 1/12 1/12 13 1 5 2 3 5 sqrt(6) 17 23 hyper_2f1(--, --, -, --------) = --------------------- 24 24 6 2 2 1/6 17 23 3125
Duh, all you need is z<1/2.