Multi-Agent Ranking framework version 2 is out

This week we released a new major version of the Multi-Agent Ranking framework (MARk). This version brings two main changes:

• there is less code to write if you want to start a project
• it has a completely new web interface with many new features

Run the demo

Easiest way to test MARk is to run the demo server, using Docker and docker-compose. Create a directory called mark and in this directory create the following docker-compose.yml :

version: '2.0'
services:
  mark-web:
    image: cylab/mark-web:1.3.2
    environment:
    - MARK_HOST=mark_mark_1
    - MARK_PORT=8080
    ports:
    - "8000:80"
    depends_on:
    - mark
  mark:
    image: cylab/mark-example:2.2.0
    environment:
    - MARK_MONGO_HOST=mark_mongo_1
    ports:
    - "8080:8080"
    depends_on:
    - mongo
  mongo:
    image: mongo:4.4

Now you can start your server with

docker-compose up

This will start 3 containers:

1. a mongodb database
2. the mark engine itself with the example configuration and detectors
3. the mark web interface

After a few seconds, the web interface will be available at http://127.0.0.1:8000. The default credentials are:

• E-Mail: mark-admin@cylab.be
• Password: change-me!

Upgrading from MARk 1.x

If you want to upgrade from MARk 1.x, here are the main changes:

• Subject and SubjectAdapter classes are not required anymore, you can remove them (also from the configuration);
• the Subject is now a Map<String, String>. In your detectors you should replace subject.getName() by subject.get("name") (for example);
• the method findData(bson.Document query) has been removed as it can be replaced by the corresponding findRawData(String label, Map<String, String> subject)
• findUniqueSubjects() and findDistinctEntries() have been removed as they were apparently not used…

This blog post is licensed under CC BY-SA 4.0