Download Example Problems - Advances Topics for Numerical Methods - Notes | CMSC 878R and more Study notes Computer Science in PDF only on Docsity! FMM CMSC 878R/AMSC 698R Lecture 18 Outline Example problem S-expansion error; S|S-translation error; S|R-translation error; R|R-translation error. Error and Neighborhoods Optimization of MLFMM within error bounds Effects of machine precision R|R-operator
Hn Hn
RaQ) = +0" = 5) — AE __prrym 2S mt "RB n()).
= mith — my!
0, m>n
CRIB rm(t) = nl poe
1
0
(RIR\E) = (AIR),,2)= | 0 0 1 3f
0
S|S-operator
(-1)"" yn!
attm — i)!
SoH) G4 are f) AD Sn),
0, m<n
ED wale) = (-1)" mn! m
eee men
nl (mt)!
1 0 0 0
t+ 100..
(SIHEQ) = m= | f ze 1 0
2 32 36 1
S|R-operator
Sn GHt) = fs t+yyrl = =f ot (1 +3 y a 7 > Ct +) em lym
ait
- re 1)"(n +n)! emt, Cy).
mint
— ("in t+}!
fl f2 ae
(SIRI) = a? 2 3.
f2 364 of.
S|S-Translation Error Translation from level α+1 to α: p first coefficients at level α can be exactly computed from p first coefficients at level α+1. Τhis is exact translation of first p coefficients! S|S-Translation Error(2) Translation from level α+1 to α: For any level α! This factor shows that we are on level α S|S-Translation Error(3) x*1 x*2 y xi In this example S|S-translations do not cause any additional error! S|R-Translation Error (3)
| Pey,xi) = PO Cy x) |
Long one!
pol pol
VY Mem - EV en
m=O n=O m=0 n=O
P-1 oo pol pol
Y Mem t PV om YY om
m=0 n=O map #=0 m=0 =O)
Pol pol Polo pol pl
Y Mem + VY emt Vem YD en
m=O x=O m=O nap map n=O m=O x=0
Pol oo pol «
Y Vem + EV on] < Den +O Dom
m=O np wep #=0 m=O nap map n=O
oo o o oo
YVemls E Vlew| = EP eml +E Ven
m=O ne, map n=O nap mm=0 map #=0
ao oo o
YD ewl+ DS Dem | = LED Kem tle)
er, mp mo mp mo continued
—>
S|R-Translation Error (4) It is really long! continued we used this S|R-Translation Error (5) Thats it! d = 1: ρ /r = 2, d = 2: ρ /r = 2(2 - √2)/√2 = 2(√2 -1) > 0.8 R|R-Translation Error(3) Indeed, in our case the regular basis functions are polynomials up to order p-1, which are obviously can be expressed via other polynomial basis up to order p-1 near arbitrary expansion center. Zero error is provided due to domains of validity are included hierarchically to larger validity domains. Total Error
AbsSingleSourceError < MaxfxpansionError(p,R,r) + AdaxSRTranstationError(p, r, p)
= (a) wept ole)
AbsTotalError = N+» AbsSingleSourceError
_ yd 2 ¢ P
-M(R) pip alos) |
since R > rtp,
AbsTotalError < ay r+? y.
Total Error(2)
d=1: p=2*, r=05-2%,
AbsTotalError < 32.
d=2: p=(2-92)24, r=0s2-2%,
QL
AbsTotalError < 3
N J2 ‘
2— /2 \ 2(/2/2+2- 2)
= 3 2N (25) < 5.2(0.6 +22.
2-72 \4- 2
Both formulae can be described as
AbsTotalError < CoP2'N = e(p,NL,d).
Total Error(5). 2-Neighborhood.
AbsTotalError < Ca?2'*N = e(p,N,L,d).
Example: V = 107, £=10,d=1:
pep 10 15 20 23
€,< 4 21077 4-107? 2-107!
Example: V = 104, £=10,d=2:
p 15 20 25 30
€,< 5+107' 2-10-7 4-10- 2-10°%
Error for Different Neighborhoods k is the size of the neighborhood, m is the cell consolidation order (in our case m = 0) Optimization of MLFMM within error bounds In the example considered, the FMM error depends on: Truncation number, p; Max level of space subdivision, L; Size of the neighborhood (1,2, or maybe other); Number of sources, N; Problem dimensionality, d. For fixed (given) N and d parameters p,L,and Neighborhood Size can be optimized. Effects of Machine Precision (1) 1.E-15 1.E-12 1.E-09 1.E-06 1.E-03 1.E-12 1.E-09 1.E-06 1.E-03 Prescribed Max Abs Error, ε M ax A bs E rr or M = N , q = q (ε), p = p (ε, l max) N=16 64 256 1024 4096 16384 Double Precision Complex Arithmetics Theoretical Error Bound Actual Max Abs Error Machine Precision Thresholds Effects of Machine Error (2) Two different computational problems: 1) Compute with a given prescribed accuracy; 2) Compute with a machine precision for a given type of float numbers. Effects of Machine Error (3) 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 Number of Source Points, N C P U T im e (s ) Straightforward Setting FFIA Data Structure FFIA Max Precision C t ti ε = 10-4 ε = 10-6 M = N , q = q (ε), p = p (ε, l max) y = ax 2 y = bx FFT MLFMM MLFMM