• 0 Vote(s) - 0 Average
• 1
• 2
• 3
• 4
• 5
 An incremental method to compute (Abel) matrix inverses bo198214 Administrator Posts: 1,391 Threads: 90 Joined: Aug 2007 07/09/2010, 06:31 AM (This post was last modified: 07/09/2010, 06:43 AM by bo198214.) I fiddled a bit around with Gottfried's suggestion of LU decomposition of the Abel matrix (though in the end the formula is independent of the LU decomposition). The annoying thing about calculating the intuitive Abel function (by solving the equation Ax=b where A is the Abel matrix, x the powerseries development of the Abel function and b=(1,0,...)) that if you want to increas the matrix size you have to solve the complete equation again without being able to use your previous solution. Now I found a way how you can compute the inverse of the matrix by using the Abel matrix. I dissect the matrix as follows, for brevity I set : means column vector and means row vector. The final incremental formula is then: Where means adding the entry 1 to the vector and is extended to a nxn matrix by filling with 0's. The deriviation is perhaps too uninteresting and cumbersome to put, but I can post it if inquired. tommy1729 Ultimate Fellow     Posts: 1,454 Threads: 350 Joined: Feb 2009 07/09/2010, 12:02 PM (This post was last modified: 07/09/2010, 12:05 PM by tommy1729.) what is an abel matrix ? i know carleman and bell matrix ... whats the difference ? maybe its the matrix equivalent of the infinite linear equations of andrews slog ... bo198214 Administrator Posts: 1,391 Threads: 90 Joined: Aug 2007 07/10/2010, 04:46 AM (This post was last modified: 07/10/2010, 04:49 AM by bo198214.) (07/09/2010, 12:02 PM)tommy1729 Wrote: what is an abel matrix ? i know carleman and bell matrix ... whats the difference ? maybe its the matrix equivalent of the infinite linear equations of andrews slog ... Oh the term was introduced by Andrew and indeed is the matrix equivalent of the infinite equation system of Andrew's slog. Its C-I transposed (C is the infinite Carleman matrix, I the infinite identity matrix) and first column removed (which consists of only 0's) and then square truncated according to your needs of precision. But the above inversion formula does not depend on A being an Abel matrix. It works for every invertible matrix. But currently I see that the internal inversion algorithms of Sage are even faster to recompute the whole inverse than doing an incremtal inverse with the above formula *sigh* Gottfried Ultimate Fellow     Posts: 786 Threads: 121 Joined: Aug 2007 07/20/2010, 12:13 PM (This post was last modified: 07/20/2010, 12:14 PM by Gottfried.) (07/09/2010, 06:31 AM)bo198214 Wrote: I fiddled a bit around with Gottfried's suggestion of LU decomposition of the Abel matrix (though in the end the formula is independent of the LU decomposition). (...)Hi Henryk - I've also tried a similar thing: to compute the new column/row of the LU-factors by the old ones - optimally by reference to the previous row/column only... without success so far. Triggered by your msg I looked at a symbolic representation, using the symbol "u" for the log of the base (which can be substituted by 1 if the base is e = exp(1)). So B is the Bell-matrix for x->exp(u*x), B1 is truncate(B - I) , Then L (lower) , D (diagonal), U (upper) the inverses of the LU-factors of B1, such if it converges, B1^-1 = U*D*L , where for the slog we need only the first column of L. Looking at the last column of U with the idea to compose it by the previous column only (or by some composition of some earlier columns) I notice, that the entries are polynomials in u and I didn't see yet a simple possibility to determine that polynomials based on that of the previous colums. The same is true analoguously for the n'th entry in D and for the n'th row in L. If I determine the coefficients for the powerseries for the slog by U*D*L[,0], then I see, that the degree of the polynomials, by which the coefficients are defined, increase binomially (n,2) and in our context of tetration I can't remember any recursionformula for such a situation. (In my treatize about the symbolic description of the coefficients for the dxp-Bell-matrix I came across the same growthrate of the order of the polynomials and did not find a formula how to compute one coefficient only by its index and possibly some known constants) It would be nice to find such a formula for "the last" column in U and "last row" in L by a short recursion - this would then also allow to dismiss that matrix-inversion(s) completely... but I think, your derivation is comparably complex (so that the inversion is even faster)? Gottfried ah - ps: I tried also to express it in terms of (u-1) or (u+1) instead of u - but with little progress. Gottfried Helms, Kassel « Next Oldest | Next Newest »

 Possibly Related Threads... Thread Author Replies Views Last Post Revisting my accelerated slog solution using Abel matrix inversion jaydfox 22 26,770 05/16/2021, 11:51 AM Last Post: Gottfried Which method is currently "the best"? MorgothV8 2 7,165 11/15/2013, 03:42 PM Last Post: MorgothV8 "Kneser"/Riemann mapping method code for *complex* bases mike3 2 9,452 08/15/2011, 03:14 PM Last Post: Gottfried Attempting to compute the kslog numerically (i.e., Kneser's construction) jaydfox 11 26,975 10/26/2009, 05:56 PM Last Post: bo198214 SAGE code for computing flow matrix for exp(z)-1 jaydfox 4 12,431 08/21/2009, 05:32 PM Last Post: jaydfox Convergence of matrix solution for base e jaydfox 6 13,365 12/18/2007, 12:14 AM Last Post: jaydfox Computing Abel function at a given center jaydfox 10 18,631 11/30/2007, 06:44 PM Last Post: andydude Matrix-method: compare use of different fixpoints Gottfried 23 36,689 11/30/2007, 05:24 PM Last Post: andydude

Users browsing this thread: 1 Guest(s) 