SCM

SCM Repository

[matrix] Diff of /pkg/TODO
ViewVC logotype

Diff of /pkg/TODO

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 2005, Wed Jul 18 14:46:20 2007 UTC revision 2151, Thu Mar 20 23:31:22 2008 UTC
# Line 1  Line 1 
1  Check for DimNames propagation in coercion and other operations.  - Check for DimNames propagation in coercion and other operations.
 ------  
 - rcond methods for sparseMatrix classes  
2    
3  - Report the problem in the Linux ldexp manual page.  The second and  - Report the problem in the Linux ldexp manual page.  The second and
4    third calls in the Synopsis should be to ldexpf and ldexpl.    third calls in the Synopsis should be to ldexpf and ldexpl.
5    
 - [,] indexing: for sparse "works", but not yet for negative indices!  
   
 - consider moving alloc3Darray from ./src/Mutils.c to  
   $(RSRC)/src/base/array.c  
 ------  
6  - provide methods for "dspMatrix" and "dppMatrix"!  - provide methods for "dspMatrix" and "dppMatrix"!
7    
8  - implement (more) methods for supporting "packed" (symmetric / triangular)  - implement (more) methods for supporting "packed" (symmetric / triangular)
# Line 18  Line 11 
11    
12    (have some dtr* <-> dtp*)    (have some dtr* <-> dtp*)
13    
 -----  
   
14  - combine the C functions for multiplication by special forms and  - combine the C functions for multiplication by special forms and
15    solution wrt special forms by using a 'right' argument and a    solution wrt special forms by using a 'right' argument and a
16    'classed' argument.    'classed' argument.
17     [done with dgeMatrix_matrix_mm();  not yet for other classes;     [done with dgeMatrix_matrix_mm();  not yet for other classes;
18      and for _crossprod()]      and for _crossprod()]
19    
 - add more comprehensive examples / tests for Schur decomposition  
   
 - arithmetic for sparse matrices:  
              <sparseMatrix>  o  <same-dim-sparseMatrix>  
   should return a sparse matrix  for at least "+" and "*" , also %%,  
   and "/" and "%/%" at least when the RHS is non-zero a scalar.  
   Challenge: nice implementation (``common non-0''; but Tsparse* is not uniq).  
   
20  -----  -----
21    
22  - "Math2" , "Math", "Arith":  - "Math2" , "Math", "Arith":
23     keep triangular and symmetric Matrices when appropriate:     keep triangular and symmetric Matrices when appropriate:
24     particularly desirable for  "Math2": round(), signif()     particularly desirable for  "Math2": round(), signif()
25    
26      For triangular matrices, more specifically make sure the four rules of
27      "triangular matrix algebra" (Golub+Van Loan 1996, 3.1.8, p.93) are
28      fulfilled; now(2008-03-06) ok for Csparse; not yet for <dtr> %*% <dtr>
29    
30  - "d" <-> "l" coercion for all "[TCR]" sparse matrices is really trivial:  - "d" <-> "l" coercion for all "[TCR]" sparse matrices is really trivial:
31    "d" -> "l" : drops the 'x' slot    "d" -> "l" : drops the 'x' slot
32    "l" -> "d" : construct an 'x' slot of all '1'    "l" -> "d" : construct an 'x' slot of all '1'
# Line 52  Line 39 
39    for all  "dsparse*" -> "lsparse*" and vice versa.    for all  "dsparse*" -> "lsparse*" and vice versa.
40    How can one do this {in a documented way} ?    How can one do this {in a documented way} ?
41    
42    - Think of constructing  setAs(...) calls automatically in order to
43      basically enable all ``sensible'' as(fromMatrix, toMatrix)  calls,
44      possibly using canCoerce(.)
45    
46    - setAs(<Mcl>,  "[dln]Matrix") for <Mcl> in {Matrix or denseMatrix + sparseMatrix}
47    
48    - When we have a packed matrix, it's a waste to go through "full" to "sparse":
49      ==> implement
50            setAs("dspMatrix", "sparseMatrix")
51            setAs("dppMatrix", "sparseMatrix")
52            setAs("dtpMatrix", "sparseMatrix")
53      and the same for "lsp" , "ltp"  and  "nsp" , "ntp" !
54    
55  - tcrossprod(x, y) : do provide methods for y != NULL  - tcrossprod(x, y) : do provide methods for y != NULL
56    calling Lapack's DGEMM for "dense"    calling Lapack's DGEMM for "dense"
57    [2005-12-xx: done for dgeMatrix at least]    [2005-12-xx: done for dgeMatrix at least]
# Line 61  Line 61 
61    str(H6 <- as(h6, "dspMatrix"))       # has lost factor    str(H6 <- as(h6, "dspMatrix"))       # has lost factor
62    ## and the same in a similar situation involving  "dpo", "dpp"    ## and the same in a similar situation involving  "dpo", "dpp"
63    
 - Things like  M[upper.tri(M)] are not really most useful for  sparse  
   matrices.  --> provide generic functions  
   upperTriMatrix(), lowerTriMatrix()  both with argument  'diag = TRUE'  
   (which can be set to FALSE of course) which are used to extract a  
   triangle from an arbitrary sparse matrix and  return a  "dtCMatrix".  
   
64  - Factorizations: LU done; also Schur()  for  *sparse*  Matrices.  - Factorizations: LU done; also Schur()  for  *sparse*  Matrices.
65    
 - band(), triu(), tril() for *all* including "matrix", not just sparse matrices  
   
66  - is.na() method for all our matrices [ ==> which(*, arr.ind=TRUE) might work ]  - is.na() method for all our matrices [ ==> which(*, arr.ind=TRUE) might work ]
67    
 - When we have a packed matrix, it's a waste to go through "full" to "sparse":  
   ==> implement  
         setAs("dspMatrix", "sparseMatrix")  
         setAs("dppMatrix", "sparseMatrix")  
         setAs("dtpMatrix", "sparseMatrix")  
   and the same for "lsp" , "ltp"  and  "nsp" , "ntp" !  
   
68  - use  .Call(Csparse_drop, M, tol) in more places,  - use  .Call(Csparse_drop, M, tol) in more places,
69    both with 'tol = 0.' to drop "values that happen to be 0" and for    both with 'tol = 0.' to drop "values that happen to be 0" and for
70    zapsmall() methods for Csparse*    zapsmall() methods for Csparse*
# Line 92  Line 77 
77  - chol() and determinant() should ``work'': proper result or "good" error  - chol() and determinant() should ``work'': proper result or "good" error
78    message.    message.
79    
 - Think of constructing  setAs(...) calls automatically in order to  
   basically enable all ``sensible'' as(fromMatrix, toMatrix)  calls,  
   possibly using canCoerce(.)  
   
80  - make sure *all* group methods have (maybe "bail-out") setMethod for "Matrix".  - make sure *all* group methods have (maybe "bail-out") setMethod for "Matrix".
81    e.g. zapsmall(<pMatrix>) fails "badly"    e.g. zapsmall(<pMatrix>) fails "badly"
82    
 - speedup: pass class definition to non0ind() [check all calls ..]  
   
83  - sum(): implement methods which work for *all* our matrices.  - sum(): implement methods which work for *all* our matrices.
84    
85  - Implement  expand(.) for the Cholesky() results  - Implement  expand(.) for the Cholesky() results
86    "dCHMsimpl" and  "dCHMsuper"  -- currently have no *decent* way to get at    "dCHMsimpl" and  "dCHMsuper"  -- currently have no *decent* way to get at
87    the matrix factors of the corresponding matrix factorization !!    the matrix factors of the corresponding matrix factorization !!
88    
89  - rbind(<sparse>, <dense>) does not work  (e.g. <dgC>, <dge>)  - rbind2(<sparse>, <dense>) does not work  (e.g. <dgC>, <dge>)
   
 - setAs(<Mcl>,  "[dln]Matrix" )  for <Mcl> in {Matrix or denseMatrix + sparseMatrix}  
   
 - Tell users about the possibility to disable the "S4-generic but somewhat slow"  
   cbind/rbind, e.g. via  
   
   setHook(packageEvent("Matrix", "onLoad"),  
           function(...) methods:::bind_activation(FALSE))  
   
 - ensure that  M[0], M[FALSE], M[1:2]  works as for traditional Matrices  
   
 - make sure  M[FALSE, FALSE]  works for all Matrices  
   {e.g. fails for M <- Diagonal(4)}  
90    
91  - <sparse> %*% <dense>  {also in crossprod/tcrossprod}  currently always  - <sparse> %*% <dense>  {also in crossprod/tcrossprod}  currently always
92    returns <dense>, since --> Csparse_dense_prod --> cholmod_sdmult    returns <dense>, since --> Csparse_dense_prod --> cholmod_sdmult
# Line 133  Line 99 
99    --> R/diagMatrix.R ('FIXME')    --> R/diagMatrix.R ('FIXME')
100    but also R/Ops.R  to ensure  sp-sym. + sp-sym. |-> sp-sym.  etc    but also R/Ops.R  to ensure  sp-sym. + sp-sym. |-> sp-sym.  etc
101    
102  - For a square sparse matrix 'b' {typically dgCMatrix or dgTMatrix},  - Diagonal(n) %*% A ---  too slow!! --> ~/R/MM/Pkg-ex/Matrix/diag-Tamas-ex.R
   we'd want a function  "Mat_plus_t_Mat" <- function(b) {....}  
   which computes the symmetric sparse matrix   b + t(b)  
   in way that never works with size-doubled vectors from  b@i etc..  
103    
104  - ! <symmetricMatrix>  loses symmetry, both for dense and sparse matrices.  - ! <symmetricMatrix>  loses symmetry, both for dense and sparse matrices.
105    !M  where M is "sparseMatrix", currently always gives dense. This only    !M  where M is "sparseMatrix", currently always gives dense. This only
106    makes sense when M is ``really sparse''.    makes sense when M is ``really sparse''.
107    
 - log1p(<sparseMatrix>) "should" give <dsparseMatrix>  
   Pretty surely, this would happen automagically, if "log1p" became part of  
   "Math" group generic ---> which is the case from R 2.6.0 on  
   
108  - msy <- as(matrix(c(2:1,1:2),2), "dsyMatrix"); str(msy)  - msy <- as(matrix(c(2:1,1:2),2), "dsyMatrix"); str(msy)
109    
110    shows that the Cholesky factorization is computed ``too quickly''.    shows that the Cholesky factorization is computed ``too quickly''.
111    Can be a big pain for largish matrices, when it is unneeded.    Can be a big pain for largish matrices, when it is unneeded.
112    
113    - example(Cholesky, echo=FALSE) ; cm <- chol(mtm); str(cm); str(mtm)
114    
115      shows that chol() does not seems to make use of an already
116      present factorization and rather uses one with more '0' in x slot.
117    
118  - diag(m) <- val    currently automatically works via  m[cbind(i,i)] <- val  - diag(m) <- val    currently automatically works via  m[cbind(i,i)] <- val
119    However,    This (`[<-` method) is now "smart" for diagonalMatrix, but needs also to
120    we need methods for 'diag<-' at least for diagonalMatrix,    be for triangularMatrix, and probably also "dense*general*Matrix" since the
   triangularMatrix, and probably also "dense*general*Matrix" since the  
121    above currently goes via "matrix" and back instead of using the 'x' slot    above currently goes via "matrix" and back instead of using the 'x' slot
122    directly.    directly; in particular, the triangular* "class property" is lost!
123    
124    - image(M, ..): Think about an optional smart option which keeps
125       "0 |-> transparent" and allows colors to differentiate negative and
126       positive entries.
127    
128    - examples for solve( Cholesky(.), b, system = c("A", "LDLt"....))
129      probably rather in man/CHMfactor-class.Rd than man/Cholesky.Rd
130    
131    - LDL(<CHMsimpl>) looks relatively easy; via  "tCsparse_diag()"
132       {diagonal entries of *triangular* Csparse}
133      --> see comment in determinant(<dsC>) in R/dsCMatrix.R, will give
134      faster determinant
135    
136    - tr(A %*% B) {and even  tr(A %*% B %*% C) ...} are also needed
137      frequently in some computations {conditional normal distr. ...}.
138      Since this can be done faster than by
139        sum(diag(A %*% B))  even for traditional matrices, e.g.
140                   sum(A * t(B)) or {even faster for "full" mat}
141                   crossprod(as.vector(A), as.vector(B))
142      and even more so for, e.g.  <sparse> %*% <dense>
143      {used in Soeren's 'gR' computations},
144      we should also provide a generic and methods.
145    
146    - qr.R(qr(x)) may differ for the "same" matrix, depending on it being
147      sparse or dense:
148        "qr.R(<sparse>) may differ from qr.R(<dense>) because of permutations"
149    
150      This is not really acceptable and currently influences  rcond() as well.
151    
152    - chol() and qr() generic:  currently have *two* arguments, and give the msg
153    
154      >  New generic for "chol" does not agree with implicit generic from package
155      >  "base"; a new generic will be assigned with package "Matrix"
156    
157      (and ditto for "qr")
158    
159      It was mentioned by an R-core member that he thought it did not make
160      sense to also dispatch on 'tol' or 'pivot' ...  --> maybe change that..
161    
162    - eigen() should become generic, and get a method at least for diagonal,
163      but also for symmetric -> dsyMatrix  [LAPACK dsyev() uses UPLO !],
164      but also simply for dgeMatrix (without going via tradition matrices).
165      What about Sparse?  There's fill-in, but it may still be sensible, e.g.
166      mlist <- list(1, 2:3, diag(x=5:3), 27, cbind(1,3:6), 100:101)
167      ee <- eigen(tcrossprod(bdiag(lapply(mlist, as.matrix))))
168      Matrix( signif(ee$vectors, 3) )
169    
170    - facmul() has no single method defined;  it looks like a good idea though
171      (instead of the infamous qr.qy, qr.qty,.... functions)
172    
173    - symmpart() and skewpart()  for *sparse* matrices still use (x +/- t(x))/2
174      and could be made more efficient.
175      Consider going via  asTuniq() or something very close to
176      .Arith.Csparse() in R/Ops.R
177    
178    - many setAs(*, "[dl]..Matrix") are still needed, as long as e.g.
179      replCmat() uses as_CspClass() and drop0(.) which itself call
180      as_CspClass() quite a bit.  --> try to replace these by
181      as(*, "CsparseMatrix"); forceSymmetric, etc.
182    
183    - add examples (and tests!) for update(<CHMfactor>, ..) and
184      Cholesky(......, Imult), also tests for hidden {hence no examples}
185      ldetL2up() { R/CHMfactor.R }

Legend:
Removed from v.2005  
changed lines
  Added in v.2151

root@r-forge.r-project.org
ViewVC Help
Powered by ViewVC 1.0.0  
Thanks to:
Vienna University of Economics and Business Powered By FusionForge