|
Size: 700
Comment:
|
← Revision 5 as of 2008-11-14 13:42:07 ⇥
Size: 860
Comment: converted to 1.6 markup
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 1: | Line 1: |
| what I did this week... looked into how to compute the RREF of big moderately sparse matrices (especially over GF(2)): | = what I did this week = |
| Line 3: | Line 3: |
| * looked into Macaulay F4 implementation, got excited/depressed about its speed at a certain point, then found out that it was buggy | looked into how to compute the RREF of big sparse to moderately dense matrices (especially over GF(2)): * learned that this is a hard problem * looked into Macaulay F4 implementation, because they seem to do sparse elimination. got excited/depressed about its speed for Cyclic14 over GF(2) at a certain point, then found out that implementation is buggy and generated incorrect results |
| Line 7: | Line 10: |
| * need more experimental data of matrices that we want to efficiently reduce: generating them | * need more experimental data of matrices that I am really interested in: generating them |
what I did this week
looked into how to compute the RREF of big sparse to moderately dense matrices (especially over GF(2)):
- learned that this is a hard problem
- looked into Macaulay F4 implementation, because they seem to do sparse elimination. got excited/depressed about its speed for Cyclic14 over GF(2) at a certain point, then found out that implementation is buggy and generated incorrect results
- paired down code for structured Gaussian elimination, writing wrappers for this
- at a certain point was not sure whether it would be easier to write wrappers or rewrite the code in Python/Cython
- compressed matrix representations (deflate() each row): didn't get very far with that, deflate() generates a significant performance impact.
- need more experimental data of matrices that I am really interested in: generating them