This repo was created for comparing three standard gossip solutions in the distributed mean problem. The question in this problem is simply that: what is the mean of some variables which are shared by a number of devices in the network? One instance could be to ask the average temperature for the bunch of sensors in a wireless network.

• Push-pull Gossip
• Push-sum Gossip
• Weighted Gossip

Please check out the enoumous number of papers if you are unfamiliar with any of them.

The repository owner tried to write a paper evaluating and comparing several accuracy and runtime efficiency of these three algorithms on three random graphs: random, scale-free, and small-world. However, due to his bad writing skills and the lack of interest to the community (incremental contribution), that paper did not work out. And later, he just turned to some other completely different topics and got locked in. Feel free to use the code.

The repo onwer would like to emphasize that the implementation is also contributed by @wuliaososhunhun. He has contributed to the codebase at least as much as I have (possibly even more than I have). He should be given the credit.

The main component in this project is the AKKA module in Java/Scala which provides a real parallel computing environment.

• Ant
• SBT
• Scala 2.11.8
• Python 2.7

One can run sbt to build the project using the following command:

sbt clean compile 

One option to run the simulations is using

sbt run 

Please note that, all the python code in the project is in Python2. Running Python3 may get bugs.

To install all the required python packages, run

pip install numpy scipy scikit-learn networkx matplotlib pandas seaborn joblib simplejson scikits.bootstrap