Re: A doubt about core.matrix
Thanks Mikera, for the detailed response. I tested the element-wise product on two sparse matrix.. it seems to run through the entire matrix ? Am I missing something. Regards, Sunil On Wed, Feb 17, 2016 at 7:47 AM, Mikera wrote: > Hi Sunil, > > You are correct that the naive code will eagerly produce the full matrix > UxV, which may be very large :-) > > I think your strategy of extracting rows and columns from U and V is the > best one. Row and column extraction for dense matrices in vectorz-clj is > very efficient since it just uses strided vectors, as is the dot product > operation, so you probably won't see much of a performance overhead from > doing it this way. I wouldn't worry about the verbosity - obviously you > should encapsulate this logic in a function if you are doing it in many > places. > > Top tip: also remember to use (non-zero-indices W) if you want to know > which elements of the sparse matrix are non-zero without iterating over > every element > > Mike. > > > On Tuesday, 16 February 2016 23:17:52 UTC+8, Sunil Nandihalli wrote: >> >> Hi Everybody, >> I am newbie to core.matrix .. I have the following expression >> >> W -> a large spare matrix of size MxN >> U -> a dense matrix of size MxK >> V -> a dense matrix of size KxN >> >> and K << (M,N) >> (require [clojure.core.matrix :as m]) >> >> I want to compute (m/mul W (m/* U V)) >> >> m/mul computes element-wise product >> >> m/* computes regular matrix multiplication >> >> I want to know if the above would compute full-matrix UV .. if it does >> then I want to know if there is an elegant way to compute the >> spare-resultant matrix without blowing up memory in the intermediate stage >> >> I am currently just extracting the corresponding rows and columns from U >> and V for the elements which are non-zero in W .. but that seems >> unnecessarily verbose... >> >> I am using the vectorz implementation of core.matrix >> >> Thanks in advance. >> >> Sunil. >> > -- > You received this message because you are subscribed to the Google > Groups "Clojure" group. > To post to this group, send email to clojure@googlegroups.com > Note that posts from new members are moderated - please be patient with > your first post. > To unsubscribe from this group, send email to > clojure+unsubscr...@googlegroups.com > For more options, visit this group at > http://groups.google.com/group/clojure?hl=en > --- > You received this message because you are subscribed to the Google Groups > "Clojure" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to clojure+unsubscr...@googlegroups.com. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
Re: A doubt about core.matrix
Hi Sunil, You are correct that the naive code will eagerly produce the full matrix UxV, which may be very large :-) I think your strategy of extracting rows and columns from U and V is the best one. Row and column extraction for dense matrices in vectorz-clj is very efficient since it just uses strided vectors, as is the dot product operation, so you probably won't see much of a performance overhead from doing it this way. I wouldn't worry about the verbosity - obviously you should encapsulate this logic in a function if you are doing it in many places. Top tip: also remember to use (non-zero-indices W) if you want to know which elements of the sparse matrix are non-zero without iterating over every element Mike. On Tuesday, 16 February 2016 23:17:52 UTC+8, Sunil Nandihalli wrote: > > Hi Everybody, > I am newbie to core.matrix .. I have the following expression > > W -> a large spare matrix of size MxN > U -> a dense matrix of size MxK > V -> a dense matrix of size KxN > > and K << (M,N) > (require [clojure.core.matrix :as m]) > > I want to compute (m/mul W (m/* U V)) > > m/mul computes element-wise product > > m/* computes regular matrix multiplication > > I want to know if the above would compute full-matrix UV .. if it does > then I want to know if there is an elegant way to compute the > spare-resultant matrix without blowing up memory in the intermediate stage > > I am currently just extracting the corresponding rows and columns from U > and V for the elements which are non-zero in W .. but that seems > unnecessarily verbose... > > I am using the vectorz implementation of core.matrix > > Thanks in advance. > > Sunil. > -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
A doubt about core.matrix
Hi Everybody, I am newbie to core.matrix .. I have the following expression W -> a large spare matrix of size MxN U -> a dense matrix of size MxK V -> a dense matrix of size KxN and K << (M,N) (require [clojure.core.matrix :as m]) I want to compute (m/mul W (m/* U V)) m/mul computes element-wise product m/* computes regular matrix multiplication I want to know if the above would compute full-matrix UV .. if it does then I want to know if there is an elegant way to compute the spare-resultant matrix without blowing up memory in the intermediate stage I am currently just extracting the corresponding rows and columns from U and V for the elements which are non-zero in W .. but that seems unnecessarily verbose... I am using the vectorz implementation of core.matrix Thanks in advance. Sunil. -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.