Healing strategies for networks

CSSS Santa Fe 2007 Home Faculty U.S. Visa Requirements Participants Schedule Readings Logistics Travel Projects & Working Groups Tutorials After Hours Photos T-shirt Design Forum Project Presentations Final Papers

Title

An investigation into healing strategies for boolean networks (?) [Perhaps lexical networks as well - see further below]

Can be seen as the boolean approach of Learning & the aging brain --Amelie

though boolean networks seems a good choice, I think we should still keep an open mind whether we want to use boolean networks only (useful for modelling brain, circuits etc where a global function is computed by the network.) or a network of intelligent nodes (like computers useful for peer-to-peer, distributed, real computer networks etc). -- Amitabh

Next Meetings

Week 3 - Thursday, 4:30pm - Great Hall (by Chris)

Week 2 - Thursday, 12:30 - Great Hall (by Chris)

Week 2 - Tuesday, 12:30 - Great Hall (by Chris)

Members

1. Alex Healing (alexhealing (at) gmail.com)
2. Amelie Veron
3. Sayan -- See suggested project on healing in lexical networks at this page: Lexical Networks
4. Christian Darabos
5. Amitabh Trehan
6. nathan menke
7. Tim Johann
8. ...

Overview

Simple

1. Random Boolean Network
2. Evolve network parameters (boolean function, connections, ...)
3. Select n best networks
4. Apply different healing strategies to each n network and compare

Evolved-Resilience Network

1. Random Boolean Network
2. Evolve with failures / insertions
3. Select n best networks
4. Apply different healing strategies

Healing Strategies

• Re-evolve whole network
• Local healing / Directed evolution
  • Static re-evolution around failure point
    • Breadth-first
    • Depth-first
  • Dynamic expansion of re-evolution

Types of failure

• Removal of node
• Removal of connections
• Byzantine failure(Insertion of 'crazy' node)

Still to decide

• Simulation environment - Java? : NetLogo?
• Stochasisity / non-determinism?
• Boolean functions, etc...

Planning and Progress

( Suggested: Please enter the aspect of the project you are working on and with whom and your suggestions for coordination)

Tim (and Chris): Have developed a boolean network in NetLogo:

• [RBN Explorer]
• [RBN Generator] - use this program to generate files for the Failure/Healing Program
• [RBN Failure/Healing Program]

Next steps: define the input nodes to the network and set of output nodes to represent a function. Suggestion at last meeting: Use the configuration of the whole network as the input. Alternate suggestion: Is there a way to have input as a fixed set of nodes so that the boolean network can itself work as a module (this would make the network analogous to the brain and the inputs analogous to the sensory neurons).

Attempt local (and other suggested healings) and study healing performance? (by Amitabh)

Background Reading

• A description of Boolean Networks with examples and pictures
• Wikipedia article with links to Stuart Kauffman's work
• File:Csusc07heal.pdf An abstract about a self-healing algorithm in reconfigurable networks -- Amitabh

...