虽然不是我写的,但我觉得很好,希望与大家分享。以下的内容转自校内:& N9 j! V5 }( ^2 J8 u
& @, a3 B& U+ ?& Q0 `从计算数学的字面来看,应该与计算机有密切的联系,也强调- o! K. U' v1 N1 @. F" {* z& c
了实践对于计算数学的重要性。也许Parlett教授的一段话能 : x8 S, j) v6 }& A最好地说明这个问题: " v N$ I' z( C $ ]9 d/ u1 V8 q* J" A( X4 v; lHow could someone as brilliant as von Neumann think7 j+ C$ N& O# y) n5 a: r8 |
hard about a subject as mundane as triangular factoriz- u$ I- k) F( b: N2 Q. _
-ation of an invertible matrix and not perceive that, 7 `+ K% P8 u, ywith suitable pivoting, the results are impressively! }; o1 D; b9 W6 e
good? Partial answers can be suggested-lack of hands-on5 x! d( }9 }# Z, A0 @; I
experience, concentration on the inverse rather than on3 ]- K( ^# z) w' F. k$ |3 b
the solution of Ax = b -but I do not find them adequate.- ~' V) S; i# M+ F2 p6 B
Why did Wilkinson keep the QR algorithm as a backup to a& A/ D2 ~3 i/ O# v
Laguerre-based method for the unsymmetric eigenproblem 1 P. Y* b- {4 U" R( y; [) ]for at least two years after the appearance of QR? Why $ c. N& e6 U' vdid more than 20 years pass before the properties of2 z" ]' |: w I, O5 a# p
the Lanczos algorithm were understood? I believe that : `+ L- X7 Q5 ithe explanation must involve the impediments to " e" O8 q2 t8 L9 ^: Rcomprehension of the effects of finite-precision 2 t2 G. `$ v6 E: T* @% {arithmetic.(引自www.siam.org/siamnews/11-03/matrix.pdf)& ^* Y; X* l* r
, s1 G) v6 k+ F" c# U* R0 r& V" U" C; S f! J" C
" t( j% H. A' C \& I' d
先订正一个错误:Sawyer的那本书的题目我4 H/ Z! y1 ~' C
记错了,应该叫《数值泛函分析初览》,系资料室和图书馆 : S: Y" ?6 y1 M9 n m都有中译本的。 9 I( T* Y9 e$ M/ O 2 C, Y1 t" E A4 g) N接下来介绍几本矩阵计算方面的书的。浙大的张振跃老师/ U( a" O' M. i0 v7 [+ Q
在这方面有很出色的工作,中科院的白中治,北京大学的徐 : G4 q% O. d! y( r+ p/ `/ C4 q! Z树方,复旦的魏益民和曹志浩,澳门大学的金小庆都是这方 , m; P, K( E4 A: A向的,还有复旦出去的柏兆俊。肯定还有许多学者在这方面! F+ j1 X; M t& C; g
有很突出的工作,可惜我基本上没什么涉略,这里也不能列 O# v; c2 d& O+ A出来。' P, y7 w! C ] d% O" P
1 w& v9 `4 }7 `' T& N
国外的大牛有Golub,很多这个方向的大家都是他的学生。 * G$ L% n- z0 [Kahan, James Demmel, Peter Stewart, L N Trefethen," t$ n6 z1 J ]9 j4 D& A
Higham,这个名单可以列的很长,这些人是矩阵计算方面 J( a$ M* z4 T
的大家。& ^5 m& c c! b
, w1 A4 d: L: Z, `. T矩阵计算方面最经典的书应该是J H Wilkinson的《The 0 W# S6 J; s: Q+ R/ ?Algebraic Eigenvalue Problem》(有中译本,石钟慈等# T* u1 i9 V' `5 _6 ^
人译,《代数特征值问题》,科学出版社,学校图书馆有,/ `! Y" p, [3 K! J
系里有英文版的)。这本书虽然老,但是据说读一下还是3 H8 V d8 a3 E; [
很有启发的。现在的经典是Golub和1 `# C; H. Q. g6 `3 E Q: E
van Loan的《Matrix Computation》(有中译本,袁亚湘译,+ j5 h, ~: q) N! o
《矩阵计算》,科学出版社),英文版的电子版可以在网上 : G& r% }: ~5 t& A找到的。其他的书有Demmel的《Applied Numerical Linear 5 h. C5 ^/ x4 ^- y+ W8 t* w6 xAlgebra》,Trefethen & Bau 的; ?2 Y, h$ s) x p3 b. s# z. U
《Numerical Linear Algebra》据说也是很好的。Yousef ' {, X2 F: z% H5 wSaad有两本书《Iterative methods for sparse systems》6 H, u: Y. k7 K* T, V
和《Numerical methods for large eigenvalue problems》,0 m' A# e" C1 K- C* T: ~- B; n7 d
写的挺有意思的,在他的主页# @' a' @$ V, {5 H; `
(http://www-users.cs.umn.edu/~saad/) 2 b: U% w( P" g& M% s& q6 c上可以down。说到矩阵计算,还得提到Householder的一本老0 E" ~6 s6 R# h; }7 \9 Z
书,《The theory of matrices in numerical analysis》 + V8 e$ a! w6 m6 g1 Q" J(有中译本,系里中英文版的都有)。, T( K. _/ m# M: ^# m
V6 M! g- a0 l, n3 \0 ILN Trefethen现在是剑桥大学的教授,他写的每一本书都很经典, 4 t) T1 I* U/ V( U! m前面已经到过他的几本书了,《Spectral Method in Matlab》, # Y; B# J2 `5 m《Numerical Linear Algebra》,还有《Finite Difference 4 b, T! m2 `; ]+ h$ U* Wand Spectral methods》(在他的主页上可以/ {. e2 n3 E- D$ k0 D. _
down,http://web.comlab.ox.ac.uk/oucl/work/nick.trefethen/) % ~6 a j3 d, }。读他的书和文章感觉也是人生的一大享受。 # d; O9 P# U1 A' h * Y5 [; s7 Z$ r1 D他在Cornell大学任教时,曾上过一门课,就是阅读数值计算的经 ; E; \' F5 L! ?5 x9 |9 L典文献。为此他写过一个短文,列举了数值计算中的十三篇经典文 ( @( ~8 \+ [6 m献,也许对大家有点启发。 6 T5 y( Z$ y0 ^9 n" L/ e* |8 [; M5 \2 p
1. Cooley & Tukey (1965) the Fast Fourier Transform8 @" U; T+ h! c8 H& F# ?- ~
2. Courant, Friedrichs & Lewy (1928) finite difference methods for PDE: l3 _) v1 e, ]6 {4 ~
3. Householder (1958) QR factorization of matrices9 k. X' _* }. h8 [
4. Curtiss & Hirschfelder (1952) stiffness of ODEs; BD formulas 1 S/ D! {% y* Y4 s1 l5. de Boor (1972) calculations with B-splines i7 q5 e. l) d% u6. Courant (1943) finite element methods for PDE + P# y5 b9 g; L) { i$ u) U7. Golub & Kahan (1965) the singular value decomposition" C$ `7 N# `+ s/ M
8. Brandt (1977) multigrid algorithms 8 o4 d2 b* w5 ?& t9. Hestenes & Stiefel (1952) the conjugate gradient iteration' K0 Y3 s4 I- G/ y; _6 ?
10. Fletcher & Powell (1963)optimization via quasi-Newton updates 5 I7 L z. g/ c11. Wanner, Hairer & Norsett (1978) order stars and applications to ODE- T! A! T' w$ U% ]5 R
12. Karmarkar (1984)interior pt. methods for linear prog.9 D1 d. G: b9 y0 J
13. Greengard & Rokhlin (1987) multipole methods for particles2 D7 C1 K M: ~# L/ F; _ R2 \( V* d
. P C7 A6 o0 i5 g他的remark也很有意思,We were struck by how young many & l( z/ \. e, u% O: vof the authors were when they wrote these **s (average8 i& j% U6 K( e! {
age: 34), and by how short an influential ** can be6 y& }! Z7 Q, h- _
(Householder: 3.3 pages, Cooley & Tukey: 4.4).这说明大家 , V) u( m# E; `; P1 _% g都还是很有希望的,呵呵。 A! l% G, W# _6 {% s7 U3 W, \+ Y4 M# l
& H1 T% f5 Y2 w5 e( {
反问题无疑是计算数学中最热门的方向之一。该方向现在有如下 6 ~5 K. K; q. `/ u4 J. V " f' I, K& [2 i几本杂志:Inverse Problems,Journal of Inverse and Ill-posed2 J1 C7 u/ e- f p
* n0 ]) `+ b' H1 \; X0 ~; U6 p
Problems, Inverse Problems in Sciences and Engineering(以前2 g$ N/ _, G; Z0 Z3 r+ z# }9 _$ y
+ M3 i, a2 w8 r4 t* ~4 d叫Inverse Problems in Engineering).第一本杂志最好,第二本杂! i; ^- Y) V; {, t n
3 i! c% s D( Y) f% O6 x
志上面有很多苏联人的工作,第三本偏向于应用。在很多高档次的2 B0 T8 J' P- |" ?1 Y5 |
" I# S i- ?+ O" d% F杂志中都有反问题方面的文章,比如SIAM Journal on Numerical- h b: s4 o$ P1 ^
! W+ @/ p: L: x7 d0 H. IAnalysis,SIAM Journal on Mathematical Analysis, SIAM0 Y% g8 r. H% ?: U' y! s: |
, w+ ^, b" r9 I" v* `4 ^ O$ I) \
Journal on Matrix Analysis and Applications,SIAM Journal on1 }+ M* R( V7 B