This is the iterative hyperedge ranking (iHypR) algorithm, which infers actors' prominence in heterogeneous networks by utilizing hyperedges' structure of objects.


Copyright:
==========
You can modify and redistribute the program. Please DO NOT remove copyright and authors information in the source code. 

For more copyright information, please refer to attached gpl.txt.

How to compile:
===============
The program was written in C++ using the Netbeans 6.9 IDE (http://www.netbeans.org) on an Opensuse 11.4 64-bit computer. You should get a compressed netbeans project folder. 

Extract the compressed folder to a convenient place, and get into 'iHypR' folder. Type 'make' command in a terminal within the folder. After these steps, you should get an executable 'ihypr'. There is a symbol link to the executable in the folder for quick access to the program.

The program was successfully compiled and ran on OpenSuse 11.4, Ubuntu 10.4, and Mac OS X 10.5.x.

How to run:
===========
In a terminal, type the executable name, following by a network and optional 0 or 1. For example:

ihypr net.txt

ihypr net.txt 1

The first command make the program to run with A-O-H relations, while the second only uses A-H relations.

The results will be stored in a text file named 'actor_n.txt', where n is an integer for number of iterations. Each line of the results is a pair of actor name and score, tab delimited.

The algorithm is very similar to HITS, however, it is not linear. Therefore, for some networks, it may not converge. In such a case, the program will stop at 1000 iterations. Usually, it takes less 100 iterations in most networks.


Input format:
=============
The program requires strict format of input file:

1) The first line must be max_actor_id max_object_id max_hyperedge_id.
2) Each of the rest lines must be actor_id object_id hyperedge_id.
3) All ids must be positive integers starting with 1.


Questions:
==========
Any concerns, please send a message to lux3@cs.rpi.edu.