hashkat.org

Quick Start

Quickstart is intended for users who want to dive in and learn as they go.

Required Programs

#k@ is compatible with both Mac OS X and Linux. Required programs:

#k@ detailed installation instructions.

Build & Configuration Files

#k@ is configured in Python then run in C++. As such, there is a build file:

hashkat/src/config_static.h

The configuration file is called:

hashkat/INFILE.yaml

INFILE.yaml contains many parameters, many of which are never changed. The simplified configuration file is:

hashkat/quickstart_INFILE.yaml

The quickstart file enables users to configure and go. It works with the default build.

  • configure quickstart_INFILE.yaml
  • save quickstart_INFILE.yaml as INFILE.yaml
  • run #k@

Don't worry about losing the original INFILE.yaml as it has the same contents as DEFAULT.yaml.

Note: Do not change DEFAULT.yaml.

What is Hashkat?

#k@ simulates an online social network such as Twitter™. In terms of graph theory agents are 'nodes' and follows 'edges'. Agents are created with certain characteristics and added to the network. Periodically, agents chose to follow other agents. Periodically, agents signal (tweet) or retransmit received signals (retweet). Follow behavior may depend on the content of signals received or characteristics of the tweeting agent.

Create Your Own Simulation

To create your own network, first configure the network for size:

  • initial_agent, max_agents, agent add_rate, max_time (simulated time) and max_real_time
  • add rate
  • the simulation will run until the smaller of max_time (simulated) or max_real_time elapses.

Configure how edges are created and destroyed:

Agents and their signals are coded based on:

  • Ideology: from 1 to 4
  • Agent_types: from 1 to 200
  • Preference_Classes (retweet rates by agent_type & tweet_type): from 1 to 2
  • Regions: EXACTLY 3
  • Languages (hardcoded): English, English+French, French, Spanish
  • Tweet_types (hardcoded) : plain, ideological, musical, humorous

Note: There must be EXACTLY 3 Regions but all but one may have add_weight = 0.

All but one of any of the above may be effectively disabled by setting its add_weight to zero. Put another way, at least one of Ideology, Agent_type, Preference_Class, Language, & Tweet_type must have an add_weight that is not zero. Advanced users may change the maximum numbers permitted in hashkat/src/config_static.h.

To see (briefly) how it comes together, below is an example of a Region configuration in quickstart_INFILE.yaml.

Sample agent_types and preference_classes are shown at bottom of this document. Further information is available in documentation.

Running Hashkat

To run #k@ the first time you must build it by typing in the command:

./build.sh

If you change the defaults in hashkat/src/config_static.h you must re-build #k@ for the changes to take effect.

Once built, you may run #k@ by typing in the command:

./run.sh

or, with a random seed generator:

./run.sh --rand

Each time you run #k@ it writes an output directory, if one already exists it will throw an error.

Therefore after your run you must remove or rename your output directory:

mv output output_my_simulation

Running Hashkat For Second Time

To run #k@ a second time, you need to first remove some files:

  • remove: hashkat/network_state.dat

Pro-tip: move network_state.dat into your output file in case you need it in future.

The commands are:

`mv network_state.dat output/network_state.dat`

To change the simulation:

  • change what is now INFILE.yaml
  • save
  • run (with either of the above commands) `

Hashkat Output

The data collected by #k@ during the simulation is sent to files stored in the output directory, created fresh with each run.

Information in #k@ is may be collected by ranks (or bins / buckets) or individually. It may be collected for the whole run or by time period (ie simulated month). Output is usually by array number not name. So languages, for example, will be output as {0,1,2,3} being {English, English+French, French, Spanish}.

Output includes data (.dat) files, which may be viewed raw with any text editor, or may be visualized.

  • Agent_Type_info.dat - one for each agent_type
  • Categories_Distro.dat
  • indegree (follows), outdegree (followed by) and culmulative degree reports for agents (each agent has own ID number)
  • degree distribution by follow model
  • statistics re follows, followbacks, hashtags, retweets
  • most popular tweet content

Files ending .gexf are configured to be visualized with Gephi.

To create a plot of your network, enter Python from the command line by typing in the command:

python

In Python, use Networkx to create a plot of your network:

import matplotlib.pyplot as plt
import networkx as nx
G = nx.read_edgelist('output/network.dat')
pos = nx.spring_layout(G,iterations=75)
nx.draw(G,pos)
plt.show()

You may analyze your network (simulation) with Networkx from Python by running /~hashkat/visualize.py:

../visualize.py

This will produce a plot of your network and save the plot to graph.svg.

See the visualization tutorial for more analysis tools.

Screenshots

Below is an example of the quickstart_INFILE.yaml agent_type configuration.

Below is an example of the quickstart_INFILE.yaml preference_class configuration.

Further Information

Implementation of #k@ is described more fully in Input documentation and in detailed tutorials with video demonstrations. A large network using 128 kb of RAM is demonstrated and other complex scenarios. For further information or feedback do not hesitate to contact us.