[Networkit] NetworKit 3.3 released

Christian Staudt christian.staudt at kit.edu
Fri Aug 22 21:57:45 CEST 2014


Announcing NetworKit 3.3

We have released version 3.3. of NetworKit, the open-source toolkit for high-performance network analysis. The package is under rapid development and so we have added several algorithmic and utility features, on top of important bug fixes.

The release will be more widely announced as soon as the website has been updated.

Best regards
Christian Staudt

christian.staudt at kit.edu
http://parco.iti.kit.edu/staudt/index-en.shtml
Institute of Theoretical Informatics - Parallel Computing Group 
Building 50.34 Room 034
Karlsruhe Institute of Technology (KIT)


Release Notes (short)

- renamed package to "networkit" according to Python packaging convention
- restructured package to enable "pip install networkit"
- improved community detection algorithms
- improved diameter algorithms
- added support for efficient, arbitrary edge attributes via edge indexing
- Eigenvector Centrality & PageRank on basis of scipy
-  spectral methods for graph partitioning  (partitioning.SpectralPartitioner), drawing  (viztools.layout.SpectralLayout) and coloring  (coloring.SpectralColoring)
- new graph generators
    - stochastic blockmodel (generators.StochasticBlockmodel)
    - Watts-Strogatz model (generators.WattsStrogatzGenerator)
    - Forest Fire model (generators.DynamicForestFireGenerator)
- union find data structure (structures/UnionFind)
- simple spanning forest algorithm (graph.SpanningForest)
- fast algorithm for partition intersection (community/PartitionIntersection)
- hub dominance in communities (community.HubDominance)
- reader for Matlab adjacency matrices
- support for reading and writing Covers
- performance improvements in Gephi streaming interface



Release Notes (long)

Fixes

- Fixed problems that could lead to incorrect NMI calculations
- Fixed node-structural Rand measure to return values apart from 0 and 1
- Connected components assigns component ids starting at 0
- Fixed PLM and ParallelConnectedComponents for graphs with deleted nodes
- Fixed edge list partition reader for firstNode != 1
- Fixed custom edge list format options in Python 3.4
- Fixed DynamicDorogovtsevMendesGenerator so that every edge, not only the first 3 edges, can be chosen
- Performance improvements in gephi streaming

New Features

- ClusteringGenerator and GraphClusteringTools are now available in Python
- New clustering generator for non-continuous clusterings
- Added support for reading and writing Covers
- Matlab object files with matrices can be read
- Partition input format extended to support the output of the Infomap implementation [0]
- BFS calculation in the Graph class supports multiple start nodes and reports the distance
- Added Watts-Strogatz generator and Forest Fire generator
- Union Find data structure
- Intersection of Partitions
- Hub dominance in partitions and covers
- Support for edge ids / edge attributes
- Faster diameter calculation both for the approximated and the exact diameter. The desired approximation ratio is now guaranteed. The approximated diameter now also supports unconnected networks, the diameter of the component with the largest diameter is returned.

API Changes

- The weighted iterator functions of the Graph class have been removed, you can now simply use a handler that accepts the weight in the normal iterator functions

[0]: http://mapequation.org/code.html











-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ira.uni-karlsruhe.de/mailman/private/networkit/attachments/20140822/5191c1bf/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 495 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <https://lists.ira.uni-karlsruhe.de/mailman/private/networkit/attachments/20140822/5191c1bf/attachment.sig>


More information about the NetworKit mailing list