login

Revision History for A369876

(Bold, blue-underlined text is an addition; faded, red-underlined text is a deletion.)

Showing entries 1-10 | older changes
Square array read by descending antidiagonals: For T(n,k), define sequence {b(i)} where b(0) = n, b(1) = k, b(i) = A000265(b(i-1) + b(i-2)). T(n,k) is the number of steps until reaching the cyclic part of {b(i)}.
(history; published version)
#45 by N. J. A. Sloane at Tue Feb 20 13:27:57 EST 2024
STATUS

proposed

approved

#44 by Yifan Xie at Tue Feb 20 09:14:26 EST 2024
STATUS

editing

proposed

#43 by Yifan Xie at Tue Feb 20 09:13:44 EST 2024
COMMENTS

We prove the cycle's existence by contradiction. Suppose {b(i)} never enters a repeated cycle, i.e., b(i) = b(i+1) = b(i+2) never holds. Since b(i) <= (b(i-1) + b(i-2))/2 for i >= 4, b(i) < max(b(i-1), b(i-2)). b(i+1) < max(b(i), b(i-1)) <= max(b(i-1), b(i-2). Thus max(b(i), b(i+1) < max(b(i-2), b(i-1)) for i >= 4, implying that max(b(2),b(3)) > max(b(4), b(5)) > ..., with a decreasing sequence of positive integers, and they can't be smaller than G. Hence this sequence can only last for (max(b(2),b(3)) - G) terms without going down to G, ending no later than max(b(2*max(b(2),b(3)) - 2*G + 1), b(2*max(b(2),b(3)) - 2*G + 2)) and the assumption never holds afterwards, meaning that the sequence enters which is a repeated cycle, so T(n,k) <= 2*max(b(3), b(4)) - 2*G + 2contradiction.

If x and y are both odd, define sequence {c(n)} where c(n) = max(b(2*n), b(2*n+1)), c(0) = max(x,y). Since 2*G | c(n) - c(n+1), the sequence can sustain (max(x,y) - G)/(2*G) terms before going down to G (then {b(n)} enters a repeated cycle), hence T(x,y) <= (max(x,y) - G)/G;

If x is even and y is odd, T(x,y) = 1 + T(y,x+y) <= (x+y)/G;

If x is odd and y is even, T(x,y) = 2 + T(x+y,x+2*y) <= (x+2*y+G)/G;

If x and y are both even, T(x,y) = 2 + T(b(2),y+b(2)) <= (y+A000265(x+y)+G)/G.

FORMULA

T(n,k) <= 2*n + 4*k - 2*gcd(n,k) + 2.

STATUS

proposed

editing

Discussion
Tue Feb 20
09:14
Yifan Xie: At least these trivial improvements should have been done.
#42 by Kevin Ryde at Mon Feb 12 02:15:03 EST 2024
STATUS

editing

proposed

Discussion
Mon Feb 12
02:19
Yifan Xie: As the cyclic part is always that form, the statements are now equivalent.
Mon Feb 19
00:59
Kevin Ryde: I believe descents show the existence of some cycle, not that it has the form b(i) = b(i+1) = b(i+2) = G.
01:03
Kevin Ryde: That upper bound of one by one steps down looks like a lot bigger than the actual value.  Something more like number of binary gcd steps?
#41 by Kevin Ryde at Mon Feb 12 02:09:38 EST 2024
COMMENTS

We prove the cycle's existence by contradiction. Suppose {b(i)} never enters a repeated cycle, i.e., b(i) = b(i+1) = b(i+2) never holds. Since b(i) <= (b(i-1) + b(i-2))/2 for i >= 5, 4, b(i) < max(b(i-1), b(i-2)). b(i+1) < max(b(i), b(i-1)) <= max(b(i-1), b(i-2). Thus max(b(i), b(i+1) < max(b(i-2), b(i-1)) for i >= 5, 4, implying that max(b(32), b(43)) > max(b(54), b(65)) > ..., with a decreasing sequence of positive integers, and they can't be smaller than G. Hence this sequence can only last for (max(b(32), b(43)) - G) terms without going down to G, ending no later than max(b(2*max(b(32), b(43)) - 2*G + 1), b(2*max(b(32), b(43)) - 2*G + 2)) and the assumption never holds afterwards, meaning that the sequence enters a repeated cycle, so T(n,k) <= 2*max(b(3), b(4)) - 2*G + 2.

Discussion
Mon Feb 12
02:15
Kevin Ryde: Statement "never enters a repeated cycle" is not equivalent to "b(i) = b(i+1) = b(i+2) never holds".  (Only after some arguments exclude other kinds of cycles.)
#40 by Kevin Ryde at Mon Feb 12 02:03:55 EST 2024
NAME

Square array read by descending antidiagonals: For T(x,yn,k), define sequence {b(ni)} where b(10) = x, n, b(21) = y, k, b(i) = A000265(b(i-1) + b(i-2)). T(x,yn,k) is the number of steps before until reaching the cyclic part of {b(ni)}.

COMMENTS

b(i) is the odd part of b(i-1) + b(i-2). Thus, so that b(i) is odd for all i >= 3. 2 and b(i) <= (b(i-1) + b(i-2))/2 for i >= 54 (i.e. where b(i-1) and b(i-2) both odd).

The cyclic part is always of the form b(i) = b(i+1) = b(i+2) means and T(n,k) = i for the place that first happens; and the sequence b enters a repeated cycle, T(x, y) = i-1. gcd(b(i), term there is b(i+1) = A000265(gcd(x, y)) is the repeating part. Denote it as n,k)) = G.

We prove the cycle's existence by contradiction. Suppose {b(ni)} never enters a repeated cycle, i.e., b(i) = b(i+1) = b(i+2) never holds. Since b(i) <= (b(i-1) + b(i-2))/2 for i >= 5, b(i) < max(b(i-1), b(i-2)). b(i+1) < max(b(i), b(i-1)) <= max(b(i-1), b(i-2). Thus max(b(i), b(i+1) < max(b(i-2), b(i-1)) for i >= 5, implying that max(b(3), b(4)) > max(b(5), b(6)) > ..., with a decreasing sequence of positive integers, and they can't be smaller than kG. Hence this sequence can only last for (max(b(3), b(4)) - kG) terms without going down to k, G, ending no later than max(b(2*max(b(3), b(4)) - 2*k G + 1), b(2*max(b(3), b(4)) - 2*k G + 2)) and the assumption never holds afterwards, meaning that the sequence enters a repeated cycle, so T(x, yn,k) <= 2*max(b(3), b(4)) - 2*k G + 2.

FORMULA

T(x,yn,k) <= 2*x n + 4*y k - 2*gcd(x, yn,k) + 2.

EXAMPLE

Square array T(x,y) begins:

Array begins:

x n\y k 1 2 3 4 5 6 7

+------------------------------

[ 1] | 0, 6, 2, 5, 3, 7, 2, ...

[ 2] | 3, 4, 5, 6, 7, 4, 6, ...

[ 3] | 1, 8, 0, 11, 4, 6, 4, ...

[ 4] | 4, 6, 5, 5, 4, 6, 10, ...

[ 5] | 3, 7, 2, 9, 0, 9, 6, ...

[ 6] | 3, 4, 3, 5, 5, 4, 6, ...

[ 7] | 1, 7, 5, 8, 3, 7, 0, ...

...

...

T(1,2) = 6 because the its sequence b begins with b(10) = 1, b(21) = 2, b(32) = A000265(1+2) = 3, b(43) = A000265(2+3) = 5, b(54) = 1, b(65) = 3, b(6) = 1, b(7) = 1, b(8) = 1, which has reached b(9i) = b(i+1, and so on) = b(i+2) at i=6.

PROG

(PARI) T(x, yn, k)={my(i=0, -1, z=0); while(z != x n || z != y, k, z=xn; xn=yk; yk=(z+xn)/2^(valuation(z+x, n, 2)); i++); i-1; };

STATUS

proposed

editing

#39 by Yifan Xie at Sun Feb 11 23:02:34 EST 2024
STATUS

editing

proposed

#38 by Yifan Xie at Sun Feb 11 23:02:24 EST 2024
EXAMPLE

x\y 1 2 3 4 5 6 7 ...

[1] 0, 6, 2, 5, 3, 7, 2, ...

#37 by Yifan Xie at Sun Feb 11 23:01:48 EST 2024
EXAMPLE

[1] 1, 7, 3, 0, 6, 4, 8, 2, 5, 3, 7, 2, ...

[2] 3, 4, 5, 6, 7, 8, 5, 7, 4, 6, ...

[3] 2, 9, 1, 12, 5, 7, 5, 8, 0, 11, 4, 6, 4, ...

[4] 5, 7, 6, 4, 6, 5, 7, 11, 5, 4, 6, 10, ...

[5] 4, 8, 3, 10, 1, 10, 7, 2, 9, 0, 9, 6, ...

[6] 3, 4, 3, 5, 5, 4, 6, 6, 5, 7, ...

[7] 2, 8, 6, 9, 4, 8, 1, 7, 5, 8, 3, 7, 0, ...

#36 by Yifan Xie at Sun Feb 11 22:57:15 EST 2024
DATA

1, 7, 4, 0, 6, 3, 5, 2, 6, 6, 9, 5, 4, 7, 1, 7, 4, 8, 8, 12, 6, 5, 5, 8, 4, 3, 5, 5, 6, 0, 6, 3, 5, 2, 16, 7, 7, 11, 5, 10, 7, 3, 2, 4, 4, 8, 10, 9, 10, 5, 7, 2, 4, 1, 15, 6, 6, 4, 9, 5, 11, 3, 7, 15, 11, 10, 6, 9, 8, 9, 4, 6, 0, 5, 5, 8, 7, 7, 3, 4, 10, 6, 14, 10, 7, 9, 5, 8, 8, 5, 4, 7, 6, 6, 2, 7, 4, 12, 5, 9, 9, 9, 7, 8, 9, 6, 4, 3, 6, 1, 6, 3, 11, 4, 8, 8, 8, 6, 14, 7, 8, 5, 5, 13, 6