Martin Maechler
2015-Mar-19 22:02 UTC
[Rd] RFC: Matrix package: Matrix products (%*%, crossprod, tcrossprod) involving "nsparseMatrix" aka sparse pattern matrices
This is a Request For Comment, also BCCed to 390 package maintainers of reverse dependencies of the Matrix package. Most users and package authors working with our 'Matrix' package will be using it for numerical computations, and so will be using "dMatrix" (d : double precision) matrix objects M, and indirectly, e.g., for M >= c will also use "lMatrix" (l: logical i.e. TRUE/FALSE/NA). All the following is **not** affecting those numerical / logical computations. A few others will know that we also have "pattern" matrices (purely binary: TRUE/FALSE, no NA) notably sparse ones, those "ngCMatrix" etc, all starting with "n" (from ``patter[n]``) which do play a prominent role in the internal sparse matrix algorithms, notably of the (underlying C code) CHOLMOD library in the so-called "symbolic" cholesky decomposition and other such operations. Another reason you may use them because they are equivalent to incidence matrices of unweighted (directed or undirected) graphs. Now, as the subject says, I'm bringing up the topic of what should happen when these matrices appear in matrix multiplications. Somewhat by design, but also partly by coincidence, the *sparse* pattern matrices multiplication in the Matrix package mostly builds on the CHOLMOD library `cholmod_ssmult()` function which implements "Boolean arithmetic" for them, instead of regular arithmetic: "+" is logical "or" "*" is logical "and". Once we map TRUE <-> 1 and FALSE <-> 0, the only difference between boolean and regular arithmetic is that "1+1 = 1" in the (mapped) boolean arithmetic, because "TRUE | TRUE" is TRUE in original logic. The drawback of using the boolean arithmetic here is the "clash" with the usual numeric arithmetic, and arithmetic in R where logical is coerced to integer (and that to "double") when certain numerical functions/operations are used. A more severe problem --- which I had not been aware of until relatively recently -- is the fact that the CHOLMD function cholmod_ssdmult(A, B) treats *both* A and B as "pattern" as soon as one of them is a (sparse) pattern matrix. And this is - I say - in clear contrast to what R users would expect: If you multiply a numeric with a "kind of logical" matrix (a pattern one), you will expect that the TRUE/FALSE matrix will be treated as a 1/0 matrix because it is combined with a numeric matrix. So we could say that in this case, the Matrix package behavior is clearly bugous .... but still it has been the behavior for the last 10 years or so. RFC 1: "Change 1": I currently propose to change this behavior for the upcoming release of Matrix (version 1.2-0), though I have no idea if dependent packages would partly fail their checks or otherwise have changed behavior subsequently. The change seems sensible, since I think if your package relied on this behavior, it was inadvertent and accidental. Still you may differ in your opinion about this change nr.1 RFC 2: "Change 2": This change would be more radical, and something I would not plan for the upcoming release of Matrix, but possibly for an update say one or two months later or so: It concerns the matrix products when *both* matrices are pattern. A situation where the boolean arithmetic may really make sense and where indeed packages may have depended on the current behavior ("T + T |--> T"). ... although that is currently only used for *sparse* pattern matrices, not for dense ones. Further, it may still seem surprising that matrix multiplication does not behave numerically for a pair of such matrices, and by the principle of "least surprise" we should provide the boolean arithmetic matrix products in another way than by the standard %*%, crossprod() and tcrossprod() functions. So one possibility could be to change the standard functions to behave numerically, and e.g., use %&% (replace the numeric "*" by a logical "&") and crossprod(A,B, boolean=TRUE), tcrossprod(A,B, boolean=TRUE) for the three boolean arithmetic version of matrix multiplications. What do you think about this? I'm particularly interested to hear from authors and users of packages such as 'arules' which IIRC explicitly work with sparse pattern matrices. Thank you for your thoughts and creative ideas, Martin Maechler, ETH Zurich
Trevor Hastie
2015-Mar-19 23:03 UTC
[Rd] RFC: Matrix package: Matrix products (%*%, crossprod, tcrossprod) involving "nsparseMatrix" aka sparse pattern matrices
Hi Martin I got stung by this last week. glmnet produces a coefficient matrix of class ?dgCMatrix? If a predictor matrix was created using sparseMatrix as follows, one gets unexpected results, as this simple example shows. My fix was easy (I always convert the predictor matrix to class ?dgCMatrix? now) Trevor> y=Matrix(diag(4)) > y4 x 4 diagonal matrix of class "ddiMatrix" [,1] [,2] [,3] [,4] [1,] 1 . . . [2,] . 1 . . [3,] . . 1 . [4,] . . . 1> z=sparseMatrix(1:4,1:4) > z4 x 4 sparse Matrix of class "ngCMatrix" [1,] | . . . [2,] . | . . [3,] . . | . [4,] . . . |> beta=as(Matrix(1:4),"dgCMatrix") > y%*%beta4 x 1 sparse Matrix of class "dgCMatrix" [1,] 1 [2,] 2 [3,] 3 [4,] 4> z%*%beta4 x 1 sparse Matrix of class "ngCMatrix" [1,] | [2,] | [3,] | [4,] |>> On Mar 19, 2015, at 3:02 PM, Martin Maechler <maechler at stat.math.ethz.ch> wrote: > > This is a Request For Comment, also BCCed to 390 package maintainers > of reverse dependencies of the Matrix package. > > Most users and package authors working with our 'Matrix' package will > be using it for numerical computations, and so will be using > "dMatrix" (d : double precision) matrix objects M, and indirectly, e.g., for > M >= c will also use "lMatrix" (l: logical i.e. TRUE/FALSE/NA). > All the following is **not** affecting those numerical / logical > computations. > > A few others will know that we also have "pattern" matrices (purely > binary: TRUE/FALSE, no NA) notably sparse ones, those "ngCMatrix" etc, > all starting with "n" (from ``patter[n]``) which do play a prominent > role in the internal sparse matrix algorithms, notably of the > (underlying C code) CHOLMOD library in the so-called "symbolic" > cholesky decomposition and other such operations. Another reason you > may use them because they are equivalent to incidence matrices of > unweighted (directed or undirected) graphs. > > Now, as the subject says, I'm bringing up the topic of what should > happen when these matrices appear in matrix multiplications. > Somewhat by design, but also partly by coincidence, the *sparse* > pattern matrices multiplication in the Matrix package mostly builds on > the CHOLMOD library `cholmod_ssmult()` function which implements > "Boolean arithmetic" for them, instead of regular arithmetic: > "+" is logical "or" > "*" is logical "and". > Once we map TRUE <-> 1 and FALSE <-> 0, the only difference between > boolean and regular arithmetic is that "1+1 = 1" in the (mapped) > boolean arithmetic, because "TRUE | TRUE" is TRUE in original logic. > > The drawback of using the boolean arithmetic here is the "clash" with > the usual numeric arithmetic, and arithmetic in R where logical is > coerced to integer (and that to "double") when certain numerical > functions/operations are used. > > A more severe problem --- which I had not been aware of until > relatively recently -- is the fact that the CHOLMD function > cholmod_ssdmult(A, B) > treats *both* A and B as "pattern" as soon as one of them is a > (sparse) pattern matrix. > And this is - I say - in clear contrast to what R users would expect: > If you multiply a numeric with a "kind of logical" matrix (a pattern > one), you will expect that the > TRUE/FALSE matrix will be treated as a 1/0 matrix because it is > combined with a numeric matrix. > So we could say that in this case, the Matrix package behavior is > clearly bugous .... but still it has been the behavior for the last 10 > years or so. > > RFC 1: "Change 1": > I currently propose to change this behavior for the upcoming release > of Matrix (version 1.2-0), though I have no idea if dependent > packages would partly fail their checks or otherwise have changed > behavior subsequently. > The change seems sensible, since I think if your package relied on > this behavior, it was inadvertent and accidental. > Still you may differ in your opinion about this change nr.1 > > RFC 2: "Change 2": > This change would be more radical, and something I would not plan for > the upcoming release of Matrix, but possibly for an update say one or > two months later or so: It concerns the matrix products when *both* > matrices are pattern. A situation where the boolean arithmetic may > really make sense and where indeed packages may have depended on the > current behavior ("T + T |--> T"). ... although that is currently > only used for *sparse* pattern matrices, not for dense ones. > > Further, it may still seem surprising that matrix multiplication does > not behave numerically for a pair of such matrices, and by the > principle of "least surprise" we should provide the boolean arithmetic > matrix products in another way than by the standard %*%, > crossprod() and tcrossprod() functions. > So one possibility could be to change the standard functions to behave > numerically, > and e.g., use %&% (replace the numeric "*" by a logical "&") and > crossprod(A,B, boolean=TRUE), tcrossprod(A,B, boolean=TRUE) > for the three boolean arithmetic version of matrix multiplications. > > What do you think about this? I'm particularly interested to hear > from authors and users of packages such as 'arules' which IIRC > explicitly work with sparse pattern matrices. > > Thank you for your thoughts and creative ideas, > Martin Maechler, ETH Zurich---------------------------------------------------------------------------------------- Trevor Hastie hastie at stanford.edu <mailto:hastie at stanford.edu> Professor, Department of Statistics, Stanford University Phone: (650) 725-2231 Fax: (650) 725-8977 URL: http://www.stanford.edu/~hastie <http://www-stat.stanford.edu/~hastie> address: room 104, Department of Statistics, Sequoia Hall 390 Serra Mall, Stanford University, CA 94305-4065 -------------------------------------------------------------------------------------- [[alternative HTML version deleted]]
Michael Hahsler
2015-Mar-20 01:15 UTC
[Rd] RFC: Matrix package: Matrix products (%*%, crossprod, tcrossprod) involving "nsparseMatrix" aka sparse pattern matrices
Hi Martin, package arules heavily relies on ngCMatrix and uses multiplication and addition for logical operations. I think it makes sense that in a mixed operation with one dgCMatrix and one ngCMatrix the ngCMatrix should be "promoted" to a dgCMatrix. The current behavior of %*% and friends is in deed confusing: > m <- matrix(sample(c(0,1), 5*5, replace=TRUE), nrow=5) > x <- as(m, "dgCMatrix") > y <- as(m, "ngCMatrix") > x %*% y 5 x 5 sparse Matrix of class "ngCMatrix" [1,] | | | . | [2,] | | | . | [3,] . . | | . [4,] . . . | . [5,] | | | | | > x %*% x 5 x 5 sparse Matrix of class "dgCMatrix" [1,] 1 2 1 . 2 [2,] 1 3 1 . 3 [3,] . . 1 2 . [4,] . . . 1 . [5,] 1 2 2 1 2 We even explicitly coerce in our code ngCMatrix to dgCMatrix to avoid this behavior. I think all these operations probably should result consistently in a dgCMatrix. I would love to see | and & for position-wise AND and OR for ngCMatrix. Thanks, -Michael On 03/19/2015 05:02 PM, Martin Maechler wrote:> This is a Request For Comment, also BCCed to 390 package maintainers > of reverse dependencies of the Matrix package. > > Most users and package authors working with our 'Matrix' package will > be using it for numerical computations, and so will be using > "dMatrix" (d : double precision) matrix objects M, and indirectly, e.g., for > M >= c will also use "lMatrix" (l: logical i.e. TRUE/FALSE/NA). > All the following is **not** affecting those numerical / logical > computations. > > A few others will know that we also have "pattern" matrices (purely > binary: TRUE/FALSE, no NA) notably sparse ones, those "ngCMatrix" etc, > all starting with "n" (from ``patter[n]``) which do play a prominent > role in the internal sparse matrix algorithms, notably of the > (underlying C code) CHOLMOD library in the so-called "symbolic" > cholesky decomposition and other such operations. Another reason you > may use them because they are equivalent to incidence matrices of > unweighted (directed or undirected) graphs. > > Now, as the subject says, I'm bringing up the topic of what should > happen when these matrices appear in matrix multiplications. > Somewhat by design, but also partly by coincidence, the *sparse* > pattern matrices multiplication in the Matrix package mostly builds on > the CHOLMOD library `cholmod_ssmult()` function which implements > "Boolean arithmetic" for them, instead of regular arithmetic: > "+" is logical "or" > "*" is logical "and". > Once we map TRUE <-> 1 and FALSE <-> 0, the only difference between > boolean and regular arithmetic is that "1+1 = 1" in the (mapped) > boolean arithmetic, because "TRUE | TRUE" is TRUE in original logic. > > The drawback of using the boolean arithmetic here is the "clash" with > the usual numeric arithmetic, and arithmetic in R where logical is > coerced to integer (and that to "double") when certain numerical > functions/operations are used. > > A more severe problem --- which I had not been aware of until > relatively recently -- is the fact that the CHOLMD function > cholmod_ssdmult(A, B) > treats *both* A and B as "pattern" as soon as one of them is a > (sparse) pattern matrix. > And this is - I say - in clear contrast to what R users would expect: > If you multiply a numeric with a "kind of logical" matrix (a pattern > one), you will expect that the > TRUE/FALSE matrix will be treated as a 1/0 matrix because it is > combined with a numeric matrix. > So we could say that in this case, the Matrix package behavior is > clearly bugous .... but still it has been the behavior for the last 10 > years or so. > > RFC 1: "Change 1": > I currently propose to change this behavior for the upcoming release > of Matrix (version 1.2-0), though I have no idea if dependent > packages would partly fail their checks or otherwise have changed > behavior subsequently. > The change seems sensible, since I think if your package relied on > this behavior, it was inadvertent and accidental. > Still you may differ in your opinion about this change nr.1 > > RFC 2: "Change 2": > This change would be more radical, and something I would not plan for > the upcoming release of Matrix, but possibly for an update say one or > two months later or so: It concerns the matrix products when *both* > matrices are pattern. A situation where the boolean arithmetic may > really make sense and where indeed packages may have depended on the > current behavior ("T + T |--> T"). ... although that is currently > only used for *sparse* pattern matrices, not for dense ones. > > Further, it may still seem surprising that matrix multiplication does > not behave numerically for a pair of such matrices, and by the > principle of "least surprise" we should provide the boolean arithmetic > matrix products in another way than by the standard %*%, > crossprod() and tcrossprod() functions. > So one possibility could be to change the standard functions to behave > numerically, > and e.g., use %&% (replace the numeric "*" by a logical "&") and > crossprod(A,B, boolean=TRUE), tcrossprod(A,B, boolean=TRUE) > for the three boolean arithmetic version of matrix multiplications. > > What do you think about this? I'm particularly interested to hear > from authors and users of packages such as 'arules' which IIRC > explicitly work with sparse pattern matrices. > > Thank you for your thoughts and creative ideas, > Martin Maechler, ETH Zurich >-- Michael Hahsler, Assistant Professor Department of Engineering Management, Information, and Systems Department of Computer Science and Engineering (by courtesy) Bobby B. Lyle School of Engineering Southern Methodist University, Dallas, Texas office: Caruth Hall, suite 337, room 311 email: mhahsler at lyle.smu.edu web: http://lyle.smu.edu/~mhahsler
Heather Turner
2015-Mar-20 08:42 UTC
[Rd] RFC: Matrix package: Matrix products (%*%, crossprod, tcrossprod) involving "nsparseMatrix" aka sparse pattern matrices
We don't use the pattern matrices, nevertheless the proposed changes sound good to me. I particularly like the suggestion to treat the matrices as numeric by default, but provide simple ways to use boolean arithmetic instead - this means that developers have access to both forms of arithmetic and it will be more obvious from the code which arithmetic is being used. Best wishes, Heather On Thu, Mar 19, 2015, at 10:02 PM, Martin Maechler wrote:> This is a Request For Comment, also BCCed to 390 package maintainers > of reverse dependencies of the Matrix package. > > Most users and package authors working with our 'Matrix' package will > be using it for numerical computations, and so will be using > "dMatrix" (d : double precision) matrix objects M, and indirectly, > e.g., for > M >= c will also use "lMatrix" (l: logical i.e. TRUE/FALSE/NA). > All the following is **not** affecting those numerical / logical > computations. > > A few others will know that we also have "pattern" matrices (purely > binary: TRUE/FALSE, no NA) notably sparse ones, those "ngCMatrix" etc, > all starting with "n" (from ``patter[n]``) which do play a prominent > role in the internal sparse matrix algorithms, notably of the > (underlying C code) CHOLMOD library in the so-called "symbolic" > cholesky decomposition and other such operations. Another reason you > may use them because they are equivalent to incidence matrices of > unweighted (directed or undirected) graphs. > > Now, as the subject says, I'm bringing up the topic of what should > happen when these matrices appear in matrix multiplications. > Somewhat by design, but also partly by coincidence, the *sparse* > pattern matrices multiplication in the Matrix package mostly builds on > the CHOLMOD library `cholmod_ssmult()` function which implements > "Boolean arithmetic" for them, instead of regular arithmetic: > "+" is logical "or" > "*" is logical "and". > Once we map TRUE <-> 1 and FALSE <-> 0, the only difference between > boolean and regular arithmetic is that "1+1 = 1" in the (mapped) > boolean arithmetic, because "TRUE | TRUE" is TRUE in original logic. > > The drawback of using the boolean arithmetic here is the "clash" with > the usual numeric arithmetic, and arithmetic in R where logical is > coerced to integer (and that to "double") when certain numerical > functions/operations are used. > > A more severe problem --- which I had not been aware of until > relatively recently -- is the fact that the CHOLMD function > cholmod_ssdmult(A, B) > treats *both* A and B as "pattern" as soon as one of them is a > (sparse) pattern matrix. > And this is - I say - in clear contrast to what R users would expect: > If you multiply a numeric with a "kind of logical" matrix (a pattern > one), you will expect that the > TRUE/FALSE matrix will be treated as a 1/0 matrix because it is > combined with a numeric matrix. > So we could say that in this case, the Matrix package behavior is > clearly bugous .... but still it has been the behavior for the last 10 > years or so. > > RFC 1: "Change 1": > I currently propose to change this behavior for the upcoming release > of Matrix (version 1.2-0), though I have no idea if dependent > packages would partly fail their checks or otherwise have changed > behavior subsequently. > The change seems sensible, since I think if your package relied on > this behavior, it was inadvertent and accidental. > Still you may differ in your opinion about this change nr.1 > > RFC 2: "Change 2": > This change would be more radical, and something I would not plan for > the upcoming release of Matrix, but possibly for an update say one or > two months later or so: It concerns the matrix products when *both* > matrices are pattern. A situation where the boolean arithmetic may > really make sense and where indeed packages may have depended on the > current behavior ("T + T |--> T"). ... although that is currently > only used for *sparse* pattern matrices, not for dense ones. > > Further, it may still seem surprising that matrix multiplication does > not behave numerically for a pair of such matrices, and by the > principle of "least surprise" we should provide the boolean arithmetic > matrix products in another way than by the standard %*%, > crossprod() and tcrossprod() functions. > So one possibility could be to change the standard functions to behave > numerically, > and e.g., use %&% (replace the numeric "*" by a logical "&") and > crossprod(A,B, boolean=TRUE), tcrossprod(A,B, boolean=TRUE) > for the three boolean arithmetic version of matrix multiplications. > > What do you think about this? I'm particularly interested to hear > from authors and users of packages such as 'arules' which IIRC > explicitly work with sparse pattern matrices. > > Thank you for your thoughts and creative ideas, > Martin Maechler, ETH Zurich
Martin Maechler
2015-Mar-20 09:33 UTC
[Rd] RFC: Matrix package: Matrix products (%*%, crossprod, tcrossprod) involving "nsparseMatrix" aka sparse pattern matrices
>>>>> Trevor Hastie <hastie at stanford.edu> >>>>> on Thu, 19 Mar 2015 16:03:38 -0700 writes:> Hi Martin > I got stung by this last week. > glmnet produces a coefficient matrix of class ?dgCMatrix? > If a predictor matrix was created using sparseMatrix as follows, > one gets unexpected results, as this simple example shows. > My fix was easy (I always convert the predictor matrix to class ?dgCMatrix? now) > Trevor >> y=Matrix(diag(4)) Considerably faster (for larger n): Diagonal(4) if you want a sparse matrix directly, there are .sparseDiagonal() and .symDiagonal() function >> y > 4 x 4 diagonal matrix of class "ddiMatrix" > [,1] [,2] [,3] [,4] > [1,] 1 . . . > [2,] . 1 . . > [3,] . . 1 . > [4,] . . . 1 there's no problem with 'y' which is a "diagonalMatrix" and only needs O(n) storage rather than diag(n), right ? >> z=sparseMatrix(1:4,1:4) >> z > 4 x 4 sparse Matrix of class "ngCMatrix" > [1,] | . . . > [2,] . | . . > [3,] . . | . > [4,] . . . | >> beta=as(Matrix(1:4),"dgCMatrix") >> y%*%beta > 4 x 1 sparse Matrix of class "dgCMatrix" > [1,] 1 > [2,] 2 > [3,] 3 > [4,] 4 >> z%*%beta > 4 x 1 sparse Matrix of class "ngCMatrix" > [1,] | > [2,] | > [3,] | > [4,] | >> Yes, the last one is what I consieder bogous. Thank you, Trevor, for the feedback! Martin >> On Mar 19, 2015, at 3:02 PM, Martin Maechler <maechler at stat.math.ethz.ch> wrote: >> >> This is a Request For Comment, also BCCed to 390 package maintainers >> of reverse dependencies of the Matrix package. >> >> Most users and package authors working with our 'Matrix' package will >> be using it for numerical computations, and so will be using >> "dMatrix" (d : double precision) matrix objects M, and indirectly, e.g., for >> M >= c will also use "lMatrix" (l: logical i.e. TRUE/FALSE/NA). >> All the following is **not** affecting those numerical / logical >> computations. >> >> A few others will know that we also have "pattern" matrices (purely >> binary: TRUE/FALSE, no NA) notably sparse ones, those "ngCMatrix" etc, >> all starting with "n" (from ``patter[n]``) which do play a prominent >> role in the internal sparse matrix algorithms, notably of the >> (underlying C code) CHOLMOD library in the so-called "symbolic" >> cholesky decomposition and other such operations. Another reason you >> may use them because they are equivalent to incidence matrices of >> unweighted (directed or undirected) graphs. >> >> Now, as the subject says, I'm bringing up the topic of what should >> happen when these matrices appear in matrix multiplications. >> Somewhat by design, but also partly by coincidence, the *sparse* >> pattern matrices multiplication in the Matrix package mostly builds on >> the CHOLMOD library `cholmod_ssmult()` function which implements >> "Boolean arithmetic" for them, instead of regular arithmetic: >> "+" is logical "or" >> "*" is logical "and". >> Once we map TRUE <-> 1 and FALSE <-> 0, the only difference between >> boolean and regular arithmetic is that "1+1 = 1" in the (mapped) >> boolean arithmetic, because "TRUE | TRUE" is TRUE in original logic. >> >> The drawback of using the boolean arithmetic here is the "clash" with >> the usual numeric arithmetic, and arithmetic in R where logical is >> coerced to integer (and that to "double") when certain numerical >> functions/operations are used. >> >> A more severe problem --- which I had not been aware of until >> relatively recently -- is the fact that the CHOLMD function >> cholmod_ssdmult(A, B) >> treats *both* A and B as "pattern" as soon as one of them is a >> (sparse) pattern matrix. >> And this is - I say - in clear contrast to what R users would expect: >> If you multiply a numeric with a "kind of logical" matrix (a pattern >> one), you will expect that the >> TRUE/FALSE matrix will be treated as a 1/0 matrix because it is >> combined with a numeric matrix. >> So we could say that in this case, the Matrix package behavior is >> clearly bugous .... but still it has been the behavior for the last 10 >> years or so. >> >> RFC 1: "Change 1": >> I currently propose to change this behavior for the upcoming release >> of Matrix (version 1.2-0), though I have no idea if dependent >> packages would partly fail their checks or otherwise have changed >> behavior subsequently. >> The change seems sensible, since I think if your package relied on >> this behavior, it was inadvertent and accidental. >> Still you may differ in your opinion about this change nr.1 >> >> RFC 2: "Change 2": >> This change would be more radical, and something I would not plan for >> the upcoming release of Matrix, but possibly for an update say one or >> two months later or so: It concerns the matrix products when *both* >> matrices are pattern. A situation where the boolean arithmetic may >> really make sense and where indeed packages may have depended on the >> current behavior ("T + T |--> T"). ... although that is currently >> only used for *sparse* pattern matrices, not for dense ones. >> >> Further, it may still seem surprising that matrix multiplication does >> not behave numerically for a pair of such matrices, and by the >> principle of "least surprise" we should provide the boolean arithmetic >> matrix products in another way than by the standard %*%, >> crossprod() and tcrossprod() functions. >> So one possibility could be to change the standard functions to behave >> numerically, >> and e.g., use %&% (replace the numeric "*" by a logical "&") and >> crossprod(A,B, boolean=TRUE), tcrossprod(A,B, boolean=TRUE) >> for the three boolean arithmetic version of matrix multiplications. >> >> What do you think about this? I'm particularly interested to hear >> from authors and users of packages such as 'arules' which IIRC >> explicitly work with sparse pattern matrices. >> >> Thank you for your thoughts and creative ideas, >> Martin Maechler, ETH Zurich > ---------------------------------------------------------------------------------------- > Trevor Hastie hastie at stanford.edu <mailto:hastie at stanford.edu> > Professor, Department of Statistics, Stanford University > Phone: (650) 725-2231 Fax: (650) 725-8977 > URL: http://www.stanford.edu/~hastie <http://www-stat.stanford.edu/~hastie> > address: room 104, Department of Statistics, Sequoia Hall > 390 Serra Mall, Stanford University, CA 94305-4065 > --------------------------------------------------------------------------------------
Martin Maechler
2015-Mar-20 10:07 UTC
[Rd] RFC: Matrix package: Matrix products (%*%, crossprod, tcrossprod) involving "nsparseMatrix" aka sparse pattern matrices
>>>>> "MH" == Michael Hahsler <mhahsler at lyle.smu.edu> >>>>> on Thu, 19 Mar 2015 20:15:37 -0500 writes:MH> Hi Martin, MH> package arules heavily relies on ngCMatrix and uses multiplication and MH> addition for logical operations. I think it makes sense that in a mixed MH> operation with one dgCMatrix and one ngCMatrix the ngCMatrix should be MH> "promoted" to a dgCMatrix. MH> The current behavior of %*% and friends is in deed confusing: >> m <- matrix(sample(c(0,1), 5*5, replace=TRUE), nrow=5) >> x <- as(m, "dgCMatrix") >> y <- as(m, "ngCMatrix") >> x %*% y MH> 5 x 5 sparse Matrix of class "ngCMatrix" MH> [1,] | | | . | MH> [2,] | | | . | MH> [3,] . . | | . MH> [4,] . . . | . MH> [5,] | | | | | >> x %*% x MH> 5 x 5 sparse Matrix of class "dgCMatrix" MH> [1,] 1 2 1 . 2 MH> [2,] 1 3 1 . 3 MH> [3,] . . 1 2 . MH> [4,] . . . 1 . MH> [5,] 1 2 2 1 2 Indeed, that is not what one should expect. MH> We even explicitly coerce in our code ngCMatrix to dgCMatrix to avoid MH> this behavior. I think all these operations probably should result MH> consistently in a dgCMatrix. Eventually. As I said, it *is* useful to work with boolean arithmetic in some cases here, so I do want to provide that .. hopefully entirely consistently as well in the future, but longer term not via '%*%' MH> I would love to see | and & for position-wise AND and OR for ngCMatrix. Well, why don't you look? ;-) These have worked for a long time already! (I checked a version from 2008) Thanks a lot, Michael, for your valuable feedback. Martin MH> Thanks, MH> -Michael MH> On 03/19/2015 05:02 PM, Martin Maechler wrote: >> This is a Request For Comment, also BCCed to 390 package maintainers >> of reverse dependencies of the Matrix package. >> >> Most users and package authors working with our 'Matrix' package will >> be using it for numerical computations, and so will be using >> "dMatrix" (d : double precision) matrix objects M, and indirectly, e.g., for >> M >= c will also use "lMatrix" (l: logical i.e. TRUE/FALSE/NA). >> All the following is **not** affecting those numerical / logical >> computations. >> >> A few others will know that we also have "pattern" matrices (purely >> binary: TRUE/FALSE, no NA) notably sparse ones, those "ngCMatrix" etc, >> all starting with "n" (from ``patter[n]``) which do play a prominent >> role in the internal sparse matrix algorithms, notably of the >> (underlying C code) CHOLMOD library in the so-called "symbolic" >> cholesky decomposition and other such operations. Another reason you >> may use them because they are equivalent to incidence matrices of >> unweighted (directed or undirected) graphs. >> >> Now, as the subject says, I'm bringing up the topic of what should >> happen when these matrices appear in matrix multiplications. >> Somewhat by design, but also partly by coincidence, the *sparse* >> pattern matrices multiplication in the Matrix package mostly builds on >> the CHOLMOD library `cholmod_ssmult()` function which implements >> "Boolean arithmetic" for them, instead of regular arithmetic: >> "+" is logical "or" >> "*" is logical "and". >> Once we map TRUE <-> 1 and FALSE <-> 0, the only difference between >> boolean and regular arithmetic is that "1+1 = 1" in the (mapped) >> boolean arithmetic, because "TRUE | TRUE" is TRUE in original logic. >> >> The drawback of using the boolean arithmetic here is the "clash" with >> the usual numeric arithmetic, and arithmetic in R where logical is >> coerced to integer (and that to "double") when certain numerical >> functions/operations are used. >> >> A more severe problem --- which I had not been aware of until >> relatively recently -- is the fact that the CHOLMD function >> cholmod_ssdmult(A, B) >> treats *both* A and B as "pattern" as soon as one of them is a >> (sparse) pattern matrix. >> And this is - I say - in clear contrast to what R users would expect: >> If you multiply a numeric with a "kind of logical" matrix (a pattern >> one), you will expect that the >> TRUE/FALSE matrix will be treated as a 1/0 matrix because it is >> combined with a numeric matrix. >> So we could say that in this case, the Matrix package behavior is >> clearly bugous .... but still it has been the behavior for the last 10 >> years or so. >> >> RFC 1: "Change 1": >> I currently propose to change this behavior for the upcoming release >> of Matrix (version 1.2-0), though I have no idea if dependent >> packages would partly fail their checks or otherwise have changed >> behavior subsequently. >> The change seems sensible, since I think if your package relied on >> this behavior, it was inadvertent and accidental. >> Still you may differ in your opinion about this change nr.1 >> >> RFC 2: "Change 2": >> This change would be more radical, and something I would not plan for >> the upcoming release of Matrix, but possibly for an update say one or >> two months later or so: It concerns the matrix products when *both* >> matrices are pattern. A situation where the boolean arithmetic may >> really make sense and where indeed packages may have depended on the >> current behavior ("T + T |--> T"). ... although that is currently >> only used for *sparse* pattern matrices, not for dense ones. >> >> Further, it may still seem surprising that matrix multiplication does >> not behave numerically for a pair of such matrices, and by the >> principle of "least surprise" we should provide the boolean arithmetic >> matrix products in another way than by the standard %*%, >> crossprod() and tcrossprod() functions. >> So one possibility could be to change the standard functions to behave >> numerically, >> and e.g., use %&% (replace the numeric "*" by a logical "&") and >> crossprod(A,B, boolean=TRUE), tcrossprod(A,B, boolean=TRUE) >> for the three boolean arithmetic version of matrix multiplications. >> >> What do you think about this? I'm particularly interested to hear >> from authors and users of packages such as 'arules' which IIRC >> explicitly work with sparse pattern matrices. >> >> Thank you for your thoughts and creative ideas, >> Martin Maechler, ETH Zurich >> MH> -- MH> Michael Hahsler, Assistant Professor MH> Department of Engineering Management, Information, and Systems MH> Department of Computer Science and Engineering (by courtesy) MH> Bobby B. Lyle School of Engineering MH> Southern Methodist University, Dallas, Texas MH> office: Caruth Hall, suite 337, room 311 MH> email: mhahsler at lyle.smu.edu MH> web: http://lyle.smu.edu/~mhahsler
Dr. Peter Ruckdeschel
2015-Mar-20 19:04 UTC
[Rd] RFC: Matrix package: Matrix products (%*%, crossprod, tcrossprod) involving "nsparseMatrix" aka sparse pattern matrices
Hi Martin, many thanks to you and Doug for providing the Matrix package in the first place, and, second, for taking us into this decision. I have only some minor comments to make: + wherever there is a usual function call involved, using an argument "boolean" as you proposed seems perfect to me + default behaviour and default values in function arguments should, even if bugous, stick to the old behaviour for backward compatibility right now, but you might still want to change this after a long enough announcement period + when it comes to arithmetic symbols, something like %&% certainly is nice to have, but the inadvertent user (like me, probably) would not know of this, unless this is documented at a prominent place + although this is against the functional paradigm of R, I would --exceptionally-- opt for a global option to change the behaviour (a) in function argument defaults and (b), more importantly, in binary arithmetic operators like %*%, *, + --- this way everybody can have the Matrix flavour he likes just my 2c, best regards, Peter Am 19.03.2015 um 23:02 schrieb Martin Maechler:> This is a Request For Comment, also BCCed to 390 package maintainers > of reverse dependencies of the Matrix package. > > Most users and package authors working with our 'Matrix' package will > be using it for numerical computations, and so will be using > "dMatrix" (d : double precision) matrix objects M, and indirectly, e.g., for > M >= c will also use "lMatrix" (l: logical i.e. TRUE/FALSE/NA). > All the following is **not** affecting those numerical / logical > computations. > > A few others will know that we also have "pattern" matrices (purely > binary: TRUE/FALSE, no NA) notably sparse ones, those "ngCMatrix" etc, > all starting with "n" (from ``patter[n]``) which do play a prominent > role in the internal sparse matrix algorithms, notably of the > (underlying C code) CHOLMOD library in the so-called "symbolic" > cholesky decomposition and other such operations. Another reason you > may use them because they are equivalent to incidence matrices of > unweighted (directed or undirected) graphs. > > Now, as the subject says, I'm bringing up the topic of what should > happen when these matrices appear in matrix multiplications. > Somewhat by design, but also partly by coincidence, the *sparse* > pattern matrices multiplication in the Matrix package mostly builds on > the CHOLMOD library `cholmod_ssmult()` function which implements > "Boolean arithmetic" for them, instead of regular arithmetic: > "+" is logical "or" > "*" is logical "and". > Once we map TRUE <-> 1 and FALSE <-> 0, the only difference between > boolean and regular arithmetic is that "1+1 = 1" in the (mapped) > boolean arithmetic, because "TRUE | TRUE" is TRUE in original logic. > > The drawback of using the boolean arithmetic here is the "clash" with > the usual numeric arithmetic, and arithmetic in R where logical is > coerced to integer (and that to "double") when certain numerical > functions/operations are used. > > A more severe problem --- which I had not been aware of until > relatively recently -- is the fact that the CHOLMD function > cholmod_ssdmult(A, B) > treats *both* A and B as "pattern" as soon as one of them is a > (sparse) pattern matrix. > And this is - I say - in clear contrast to what R users would expect: > If you multiply a numeric with a "kind of logical" matrix (a pattern > one), you will expect that the > TRUE/FALSE matrix will be treated as a 1/0 matrix because it is > combined with a numeric matrix. > So we could say that in this case, the Matrix package behavior is > clearly bugous .... but still it has been the behavior for the last 10 > years or so. > > RFC 1: "Change 1": > I currently propose to change this behavior for the upcoming release > of Matrix (version 1.2-0), though I have no idea if dependent > packages would partly fail their checks or otherwise have changed > behavior subsequently. > The change seems sensible, since I think if your package relied on > this behavior, it was inadvertent and accidental. > Still you may differ in your opinion about this change nr.1 > > RFC 2: "Change 2": > This change would be more radical, and something I would not plan for > the upcoming release of Matrix, but possibly for an update say one or > two months later or so: It concerns the matrix products when *both* > matrices are pattern. A situation where the boolean arithmetic may > really make sense and where indeed packages may have depended on the > current behavior ("T + T |--> T"). ... although that is currently > only used for *sparse* pattern matrices, not for dense ones. > > Further, it may still seem surprising that matrix multiplication does > not behave numerically for a pair of such matrices, and by the > principle of "least surprise" we should provide the boolean arithmetic > matrix products in another way than by the standard %*%, > crossprod() and tcrossprod() functions. > So one possibility could be to change the standard functions to behave > numerically, > and e.g., use %&% (replace the numeric "*" by a logical "&") and > crossprod(A,B, boolean=TRUE), tcrossprod(A,B, boolean=TRUE) > for the three boolean arithmetic version of matrix multiplications. > > What do you think about this? I'm particularly interested to hear > from authors and users of packages such as 'arules' which IIRC > explicitly work with sparse pattern matrices. > > Thank you for your thoughts and creative ideas, > Martin Maechler, ETH Zurich-- Dr. habil. Peter Ruckdeschel, Abteilung Finanzmathematik, F3.17 Fraunhofer ITWM, Fraunhofer Platz 1, 67663 Kaiserslautern Telefon: +49 631/31600-4699 Fax: +49 631/31600-5699 E-Mail : peter.ruckdeschel at itwm.fraunhofer.de http://www.itwm.fraunhofer.de/abteilungen/finanzmathematik/mitarbeiterinnen/mitarbeiter/dr-peter-ruckdeschel.html
Reasonably Related Threads
- RFC: Matrix package: Matrix products (%*%, crossprod, tcrossprod) involving "nsparseMatrix" aka sparse pattern matrices
- RFC: Matrix package: Matrix products (%*%, crossprod, tcrossprod) involving "nsparseMatrix" aka sparse pattern matrices
- Error from sparse pattern matrix product
- R 3.5.0 fails its regression test suite on Linux/x86_64
- using apply with sparse matrix from package Matrix