There aren't built-in data structures defined in Julia's standard library right now for higher-dimensional sparse matrices, no. But you can certainly come up with your own data structure and use it however you like. Are there any dimensions in your problem along which every 2-dimensional slice has the same sparsity pattern? If there are certain index sets you'll want to be doing reductions over more often than others, then you can pick a data structure appropriately. Or for maximum generality you could use a coordinate representation.
On Sunday, September 13, 2015 at 7:28:42 AM UTC-7, Frank Kampas wrote: > > I'm doing sequential linear programming on quadratic constraints. Using > matrices makes this much more straight-forward. Without 4th rank matrices, > I have to generate a large number of 2nd rank matrices for every iteration. > However, I gather from your answer that only 2nd rank sparse matrices can > be created. > > On Saturday, September 12, 2015 at 9:16:41 PM UTC-4, Tony Kelman wrote: >> >> In JuMP you can do indexing over constraints and variables with any >> number of indexes. You probably don't need to worry about explicitly >> forming constraint matrices at all, since the flattened individual indexes >> of optimization variables and constraints are somewhat arbitrary and will >> mostly likely be rearranged by presolve in any good optimization solver. >> Just express the variables and constraints of the problem you want to solve >> and JuMP will handle the messy transformations. >> >> >> On Saturday, September 12, 2015 at 3:15:34 PM UTC-7, Frank Kampas wrote: >>> >>> >>> >>> On Saturday, September 12, 2015 at 12:09:11 PM UTC-4, Frank Kampas wrote: >>>> >>>> Is it possible to create sparse matrices with a rank other than 2? >>>> >>> >>> I've been using 4th rank sparse matrices in Mathematica for circle >>> packing. The constraints >>> can be expressed using 2nd rank matrices and representing all the pairs >>> of circles gives me >>> another two dimensions. I'd like to move the code to Julia to take >>> advantage of the access >>> to various linear programming software packages. >>> >>
