- Jul 14, 2022
-
-
Paul Nitzke authored
Remove some leftover unused imports See merge request !45
-
Paul Nitzke authored
-
Paul Nitzke authored
Fix: Correct labeling to display column partition See merge request !44
-
Paul Nitzke authored
* Contains a full replacement for the adjacency matrix test that uses the matrix from our paper. This makes the test easily verifyable by looking at the graph in the paper
-
Paul Nitzke authored
* Before: Although we partition columns the output suggested that the row nodes were colored in bipartite graphs * This was caused by mislabeling in DotWriter, swapping c_ and r_ label prefixes. * The way the labels are currently written this commit also swaps the pseudo side flag used for adjacency graphs for correct output
-
Paul Nitzke authored
* Old terminology was inconsistent with rest of the program and most of the literature on matrices * x corresponded to i and y to j which is an uncommon mapping of the axes * This does not change how the graph nodes are created from the indices
-
- Jul 09, 2022
-
-
Paul Nitzke authored
Various matrix improvements See merge request !42
-
Paul Nitzke authored
* We only use Multiply to compute the compressed matrix. With our current implementation this is always DoubleMatrix * IntegerMatrix. * As the Multiply methods are all the same except for the signature we delete the unused ones * This can be considered temporary until we solve the matrix generics problem for good
-
Paul Nitzke authored
* Improves performance for all commands that output a matrix
-
Paul Nitzke authored
Fix: Fix incorrect dimensions for seed matrix See merge request !41
-
Paul Nitzke authored
* Seed matrix now receives the n dimension of the input matrix as its n dimension
-
Paul Nitzke authored
Fix: Change backing datastructure of graphs See merge request !40
-
Paul Nitzke authored
* Replaces List with Dictionary with the node index being the key and the GraphNode (also containing the index) the value * This dramatically speeds up the MatrixMarket string to graph transformation in MatrixMarketReader since adding edges and in particular nodes to the graph is much faster * This is because the nodes are added to the graph as needes i.e. the graph searches for them in the existing node list and if not present adds them. This search is significant the larger the graph gets. This is also why the transformation of larger graphs took exponentially longer.
-
- Jul 08, 2022
-
-
Paul Nitzke authored
Fix: Cache ColorsUsed for huge performance boost See merge request !39
-
Paul Nitzke authored
Properly add sample matrices used in demos See merge request !38
-
Paul Nitzke authored
* This should theoretically improve performance as we only calculate it once instead of once for every colored node
-
Paul Nitzke authored
* This commit provides the proper fix mentioned in cd423d9d * We still keep the caching implemented in cd423d9d
-
Paul Nitzke authored
* This greatly increases performance (Many orders of magnite faster) * ColorsUsed is implemented so that every time it is queried the value is calculated on the fly * When writing ColorsUsed I did not think this would be a problem since we only need it once: When calculating the color * But the actual implementation used it at two points so that for x colored nodes the ColorsUsed were calculated twice. * Ideally this would be fixed by changing the ColorsUsed implementation in the -Graph classes. This might be fixed later so consider this commit a hotfix
-
Paul Nitzke authored
Fix: Improve README formatting See merge request !37
-
Paul Nitzke authored
Run command is now highlighted better
-
Paul Nitzke authored
These example matrices were used in multiple presentations that accompanied this project. It includes the bipartite graph from our coloring paper as well as select matrices from the SuiteSparse collection. The SuiteSparse matrices are distributed under the CC BY 4.0 license.
-
- Jul 03, 2022
-
-
Paul Nitzke authored
Add License See merge request !36
-
Paul Nitzke authored
-
- Jul 01, 2022
-
-
Paul Nitzke authored
Fix: Correct package name for some Convert classes See merge request !35
-
Paul Nitzke authored
-
- Jun 27, 2022
-
-
Paul Nitzke authored
Delete auto-generated unit test See merge request !34
-
Paul Nitzke authored
-
Paul Nitzke authored
Make node counts attributes See merge request !33
-
Paul Nitzke authored
Same reason as for 5fd5ce0a
-
Paul Nitzke authored
* Also fix all usages
-
Paul Nitzke authored
Change all strings to Strings See merge request !32
-
Paul Nitzke authored
* Better consistency * string is just an alias for String * As Strings are objects it makes sense to refer to them with according notation
-
- Jun 26, 2022
-
-
Paul Nitzke authored
Provide instructions on how to run tests See merge request !31
-
Paul Nitzke authored
-
Paul Nitzke authored
-
Paul Nitzke authored
Change build to single file release See merge request !30
-
Paul Nitzke authored
Add tests for Convert package See merge request !29
-
Paul Nitzke authored
-
Paul Nitzke authored
-
- Jun 23, 2022
-
-
Dennis Wobbe authored
-