To run the latest versionCLICK HERE . PRESENTATION PPT
Download paper: Dynamic Learning, Herding and Guru Effects in Networks (PDF 1.2 MB)
The Herding Simulator is a simple tool written in VB6 that allows the user to set up a network, and run simulations to analyse herd behaviour, clustering and guru effects under different types of network topologies. The program is based on Andreas Krause's model, outlined in [1]. More advanced features such as Differential Memory and Dynamic Weights have been added to this model.
To run simulations, first create a network, by selecting the appropriate settings for network size & shape, and network type, and clicking the "Create Network" button. Then, on the lower part of the window, select the price function, and click on "Start Simulation". Once the simulation has finished, you can analyze the degree distribution by going into the Tools menu, and selecting Network Statistics.
What's new in version 1.11?
- dynamic animaiton of star network formation
(Click on images to enlarge)
If you have the VB libraries already installed on your PC, you can run the program by clicking on the link at the top of the page. Otherwise, you will need to download this setup, unzip the files in a temporary folder, and run the setup.exe. The installation will create an icon under Start>Program Files>Herding Simulations.
Feel free to email me any questions/bugs/suggestions to aalent (non-Essex users should add @essex.ac.uk to create full e-mail address)
Have fun with it!
Amadeo
Run verison 1.10 by CLICK HERE
What's new in version 1.10?
- added the calculation for "Herding coefficient" under Network Statistics
Run version 1.9 by Click here
What's new in version 1.9?
- fixed bug where links disappeared when setting the Minimum Threshold setting to 0
- fixed bug where the statistical calculations for the theoretical values were incorrect when having the option of "Enforcing self-inference" turned on.
Run version 1.8 by clicking here.
Run version 1.7 by clicking here.
Run version 1.6 by clicking here.
What's new:
- Added a new pricing function "Random rewards with correlation"
- Added an option to "Enforce self-inference", that is, nodes cannot break the link with themselves
- Modified the algorithm, so the weight of the self-inference link is the sum of the outgoing links
Run version 1.5 by clicking here.
What's new:
-Added a minority price function
- fixed problem with the calculation of clustering coefficient
- added the theoretical values for a random graph with the "network statistics"
Run version 1.4 by clicking here.
What's new:
- Menu functions to show network statistics
Run version 1.3 by clicking here.
What's new:
- Option of choosing Lattice/Toroid shape
- fixed some bugs on the switching of links when using Dynamic Weights
- added a scroll bar to play back the animation
- Added a Tools menu for exporting price and weights data to Excel
Run version 1.2 by clicking here.
What's new:
- doesn't round up the price change when using the P(t) function. However, the rewards are still given as +1 or -1
- added an Animation feature, to dynamically see which nodes are "Buyers" and which nodes are "Sellers"
- clicking on a node allows the user to manually change the weights of the links
- dynamic weight adjustment
Run version 1.1 by clicking here.
What's new:
- more types of networks available
- different pricing functions
- new graph to view Price movement
- ability to Export data to MS Excel for further analysis
Run version 1.0 by clicking here (first version of the program 17th Oct 2003)
To run the simulation without interaction between the nodes, and notice the randomness of the network with no interaction, you can do any of the following:
- set M = 0 (if we don't use past experience, there will not be any interactions)
- set p = 0 (if there are no connections, no interaction)
- use the option button for no interaction
[1] Krause, A. "Herding without Following the Herd: The Dynamics of Case-based Decisions with Local Interactions", University of Bath