Research showing why hierarchy exists will aid the development of artificial intelligence

Research showing why hierarchy exists will aid the development of artificial intelligence
New research explains why so many biological networks, including the human brain (a network of neurons), exhibit a hierarchical structure, and will improve attempts to create artificial intelligence. The study, published in PLOS Computational Biology, demonstrates this by showing that the evolution of hierarchy -- a simple system of ranking -- in biological networks may arise because of the costs associated with network connections. Credit: Steven T. Caputo, CereberalHack.com

New research explains why so many biological networks, including the human brain (a network of neurons), exhibit a hierarchical structure, and will improve attempts to create artificial intelligence. The study, published in PLOS Computational Biology, demonstrates this by showing that the evolution of hierarchy - a simple system of ranking - in biological networks may arise because of the costs associated with network connections.

Like large businesses, many are hierarchically organised, such as gene, protein, neural, and metabolic networks. This means they have separate units that can each be repeatedly divided into smaller and smaller subunits. For example, the has separate areas for motor control and tactile processing, and each of these areas consist of sub-regions that govern different parts of the body.

But why do so many biological networks evolve to be hierarchical? The results of this paper suggest that hierarchy evolves not because it produces more efficient networks, but instead because hierarchically wired networks have fewer connections. This is because connections in biological networks are expensive - they have to be built, housed, maintained, etc. - and there is therefore an evolutionary pressure to reduce the number of connections.

In addition to shedding light on the emergence of hierarchy across the many domains in which it appears, these findings may also accelerate future research into evolving more complex, intelligent computational brains in the fields of artificial intelligence and robotics.

Researchers from the University of Wyoming and INRIA (France) led by Henok S. Mengistu simulated the evolution of computational brain models, known as , both with and without a cost for . They found that hierarchical structures emerge much more frequently when a cost for connections is present.

Author Jeff Clune says, "For over a decade we have been on a quest to understand why networks evolve to have the properties of modularity, hierarchy, and regularity. With these results, we have now uncovered evolutionary drivers for each of these key properties." Mengistu notes: "The findings not only explain why biological networks are hierarchical, they might also give an explanation for why many man-made systems such as the Internet and road systems are also hierarchical."

Author Joost Huizinga adds "The next step is to harness and combine this knowledge to evolve large-scale, structurally organized networks in the hopes of creating better and increasing our understanding of the evolution of animal intelligence, including our own."


Explore further

Researchers solve biological mystery and boost artificial intelligence

Journal information: PLoS Computational Biology

Citation: Research showing why hierarchy exists will aid the development of artificial intelligence (2016, June 9) retrieved 19 October 2019 from https://phys.org/news/2016-06-hierarchy-aid-artificial-intelligence.html
This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.
77 shares

Feedback to editors

User comments

Jun 09, 2016
IMHO processing itself is largely a matter of entropy reduction - eliminating redundancy, precisely for thermodynamic reasons, almost inadvertently also results in processing logic, and thus deriving solutions.

Optimal connective entropies also imply correspondingly-optimum response bandwidths of those networks. A dynamically self-organising network has a baseline thermo-geometric 'ground state' - a minimal connnective entropy below which it fragments into disparate sub-networks.

With respect to such a heirarchical processing system, information itself is kind of intrinsically pre-formated - there's a minimal, elementary level of complexity or order for any particular form of information in relation to its respective processing nuclei.

It is this principle that we are experiencing when we perceive the seemingly-paradoxical equivalence of octaves. Replicating it will be a milestone towards AI capable of processing information in fundamentally the same way as animals..

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more