Talk:Fast folding algorithm - Revision history

Qwerfjkl (bot): Implementing WP:PIQA (Task 26)

2024-01-31T03:52:37Z

← Previous revision		Revision as of 03:52, 31 January 2024
Line 1:		Line 1:
	{{WikiProject ~~Computing~~\|class=Stub\|~~importance=\|auto=yes}}~~		{{WikiProject banner shell\|class=Stub\|
			{{WikiProject Computing\|importance=\|auto=yes}}
			}}
	I can't find a good description of this algorithm on the net, so I've started my own. You'll notice it's got a big hole where all the technical detail should be! Contributions welcome ...		I can't find a good description of this algorithm on the net, so I've started my own. You'll notice it's got a big hole where all the technical detail should be! Contributions welcome ...

2001:CC0:202C:1001:515D:3B9B:621A:E61C: /* How it works */

2015-06-08T15:27:03Z

How it works

← Previous revision		Revision as of 15:27, 8 June 2015
Line 11:		Line 11:


	''		'''
	could I know the contact method of the author? I have several ffa questions to ask. Thanks.''		could I know the contact method of the author? I have several ffa questions to ask. Thanks.'''

2001:CC0:202C:1001:515D:3B9B:621A:E61C: /* How it works */

2015-06-08T15:26:42Z

How it works

← Previous revision		Revision as of 15:26, 8 June 2015
Line 10:		Line 10:
	In a way, this is very much like an FFT algorithm where the complex values have been replaced by profiles and the multiplication by "twiddle factors", roots of unity, has been replaced by cyclic shifts. <span style="font-size: smaller;" class="autosigned">—Preceding [[Wikipedia:Signatures\|unsigned]] comment added by [[Special:Contributions/65.95.19.27\|65.95.19.27]] ([[User talk:65.95.19.27\|talk]]) 07:33, 21 March 2010 (UTC)</span><!-- Template:UnsignedIP --> <!--Autosigned by SineBot-->		In a way, this is very much like an FFT algorithm where the complex values have been replaced by profiles and the multiplication by "twiddle factors", roots of unity, has been replaced by cyclic shifts. <span style="font-size: smaller;" class="autosigned">—Preceding [[Wikipedia:Signatures\|unsigned]] comment added by [[Special:Contributions/65.95.19.27\|65.95.19.27]] ([[User talk:65.95.19.27\|talk]]) 07:33, 21 March 2010 (UTC)</span><!-- Template:UnsignedIP --> <!--Autosigned by SineBot-->


⚫	could I know the contact method of the author? I have several ffa questions to ask. Thanks.
			''
		⚫	could I know the contact method of the author? I have several ffa questions to ask. Thanks.''

2001:CC0:202C:1001:515D:3B9B:621A:E61C: /* How it works */

2015-06-08T15:25:40Z

How it works

← Previous revision		Revision as of 15:25, 8 June 2015
Line 9:		Line 9:

	In a way, this is very much like an FFT algorithm where the complex values have been replaced by profiles and the multiplication by "twiddle factors", roots of unity, has been replaced by cyclic shifts. <span style="font-size: smaller;" class="autosigned">—Preceding [[Wikipedia:Signatures\|unsigned]] comment added by [[Special:Contributions/65.95.19.27\|65.95.19.27]] ([[User talk:65.95.19.27\|talk]]) 07:33, 21 March 2010 (UTC)</span><!-- Template:UnsignedIP --> <!--Autosigned by SineBot-->		In a way, this is very much like an FFT algorithm where the complex values have been replaced by profiles and the multiplication by "twiddle factors", roots of unity, has been replaced by cyclic shifts. <span style="font-size: smaller;" class="autosigned">—Preceding [[Wikipedia:Signatures\|unsigned]] comment added by [[Special:Contributions/65.95.19.27\|65.95.19.27]] ([[User talk:65.95.19.27\|talk]]) 07:33, 21 March 2010 (UTC)</span><!-- Template:UnsignedIP --> <!--Autosigned by SineBot-->

			could I know the contact method of the author? I have several ffa questions to ask. Thanks.

SineBot: Signing comment by 65.95.19.27 - "/* How it works */ new section"

2010-03-21T07:33:45Z

Signing comment by 65.95.19.27 - "How it works: new section"

← Previous revision		Revision as of 07:33, 21 March 2010
Line 8:		Line 8:
	During one pass of the FFA, all periods between n samples and n+1 samples are folded at once. The key idea is to divide the time series into m nonoverlapping blocks of n samples. Then the period-n profile is obtained by simply adding all the blocks. For a period between n and n+1, if one were simply folding the data, one would accumulate the profiles, but because the period is slightly longer than n, the phase would drift slightly. When the phase was wrong by one bin, you would begin cyclically shifting profiles to compensate for the drift. Thus to look at all possible periods, you should look at many different sequences of cyclic shifts. But the set of folds of m periods can be obtained from the set of folds of m/2 periods on the first and second halves by combining them both with and without shifting. So there's a divide-and-conquer algorithm.		During one pass of the FFA, all periods between n samples and n+1 samples are folded at once. The key idea is to divide the time series into m nonoverlapping blocks of n samples. Then the period-n profile is obtained by simply adding all the blocks. For a period between n and n+1, if one were simply folding the data, one would accumulate the profiles, but because the period is slightly longer than n, the phase would drift slightly. When the phase was wrong by one bin, you would begin cyclically shifting profiles to compensate for the drift. Thus to look at all possible periods, you should look at many different sequences of cyclic shifts. But the set of folds of m periods can be obtained from the set of folds of m/2 periods on the first and second halves by combining them both with and without shifting. So there's a divide-and-conquer algorithm.

	In a way, this is very much like an FFT algorithm where the complex values have been replaced by profiles and the multiplication by "twiddle factors", roots of unity, has been replaced by cyclic shifts.		In a way, this is very much like an FFT algorithm where the complex values have been replaced by profiles and the multiplication by "twiddle factors", roots of unity, has been replaced by cyclic shifts. <span style="font-size: smaller;" class="autosigned">—Preceding [[Wikipedia:Signatures\|unsigned]] comment added by [[Special:Contributions/65.95.19.27\|65.95.19.27]] ([[User talk:65.95.19.27\|talk]]) 07:33, 21 March 2010 (UTC)</span><!-- Template:UnsignedIP --> <!--Autosigned by SineBot-->

65.95.19.27: /* How it works */ new section

2010-03-21T07:33:18Z

How it works: new section

← Previous revision		Revision as of 07:33, 21 March 2010
Line 1:		Line 1:
	{{WikiProject Computing\|class=Stub\|importance=\|auto=yes}}		{{WikiProject Computing\|class=Stub\|importance=\|auto=yes}}
	I can't find a good description of this algorithm on the net, so I've started my own. You'll notice it's got a big hole where all the technical detail should be! Contributions welcome ...		I can't find a good description of this algorithm on the net, so I've started my own. You'll notice it's got a big hole where all the technical detail should be! Contributions welcome ...

			== How it works ==

			I implemented an FFA in python some time ago, so I can give a rough description of how it works.

			During one pass of the FFA, all periods between n samples and n+1 samples are folded at once. The key idea is to divide the time series into m nonoverlapping blocks of n samples. Then the period-n profile is obtained by simply adding all the blocks. For a period between n and n+1, if one were simply folding the data, one would accumulate the profiles, but because the period is slightly longer than n, the phase would drift slightly. When the phase was wrong by one bin, you would begin cyclically shifting profiles to compensate for the drift. Thus to look at all possible periods, you should look at many different sequences of cyclic shifts. But the set of folds of m periods can be obtained from the set of folds of m/2 periods on the first and second halves by combining them both with and without shifting. So there's a divide-and-conquer algorithm.

			In a way, this is very much like an FFT algorithm where the complex values have been replaced by profiles and the multiplication by "twiddle factors", roots of unity, has been replaced by cyclic shifts.

Michael Hardy: moved Talk:Fast Folding Algorithm to Talk:Fast folding algorithm: WP:MOSMATH

2010-02-28T01:07:42Z

moved Talk:Fast Folding Algorithm to Talk:Fast folding algorithm: WP:MOSMATH

← Previous revision	Revision as of 01:07, 28 February 2010
(No difference)

TinucherianBot: Autoassessment for WP:COMP : ( FAQ ) : (Plugin++) class=Stub, auto=yes.

2008-09-26T12:50:06Z

Autoassessment for WP:COMP : ( FAQ ) : (Plugin++) class=Stub, auto=yes.

← Previous revision		Revision as of 12:50, 26 September 2008
Line 1:		Line 1:
	{{WikiProject Computing\|class=\|importance=}}		{{WikiProject Computing\|class=Stub\|importance=\|auto=yes}}
	I can't find a good description of this algorithm on the net, so I've started my own. You'll notice it's got a big hole where all the technical detail should be! Contributions welcome ...		I can't find a good description of this algorithm on the net, so I've started my own. You'll notice it's got a big hole where all the technical detail should be! Contributions welcome ...

TinucherianBot: WP:COMP Tagging ! (False Postive ?? ) :(Plugin++) Added {{WikiProject Computing}}.

2008-06-25T11:11:54Z

WP:COMP Tagging ! (False Postive ?? ) :(Plugin++) Added {{WikiProject Computing}}.

← Previous revision		Revision as of 11:11, 25 June 2008
Line 1:		Line 1:
			{{WikiProject Computing\|class=\|importance=}}
	I can't find a good description of this algorithm on the net, so I've started my own. You'll notice it's got a big hole where all the technical detail should be! Contributions welcome ...		I can't find a good description of this algorithm on the net, so I've started my own. You'll notice it's got a big hole where all the technical detail should be! Contributions welcome ...

Tregoweth: moved Talk:Fast folding algorithm to Talk:Fast Folding Algorithm

2006-01-23T23:05:44Z

moved Talk:Fast folding algorithm to Talk:Fast Folding Algorithm

← Previous revision	Revision as of 23:05, 23 January 2006
(No difference)