Novel Substrates and Models for the Emergence of
Developmental, Learning and Cognitive Capabilities
9th Joint IEEE International Conference on Development and Learning and on Epigenetic Robotics (IEEE ICDL-EPIROB 2019), Oslo, Norway
Workshop Date: August 19th, 2019 - Engineer's House Conference Centre, Oslo, Norway
Paper submission deadline: May 1st, 2019
Stefano Nichele, Oslo Metropolitan University, Oslo, Norway
Jianhua Zhang, Oslo Metropolitan University, Oslo, Norway
Scope and call for papers
ICDL-EpiRob is a unique conference gathering researchers from the domains of computer science, robotics, psychology and developmental studies to share knowledge and new research results on how humans and animals develop sensing, perception, reasoning, and actions.
In order for sensing, reasoning, development and learning to take place in artificial systems/beings, computing substrates and models that support the emergence of such properties are required.
This workshop encompasses the understanding, analysis, modeling and development of novel substrates that show development and learning abilities, such as (but not limited to):
- Cellular automata and random boolean networks,
- Reservoir computing,
- Novel nanoscale materials (e.g. carbon nanotubes),
- Biological materials and systems (e.g. neuronal cultures),
- Evolvable hardware and neuromorphic computing systems,
- Novel neural models and models of biological neuron,
- Evolution-in-materio systems,
- Computational matter,
- Slime mould computing,
- Micro- and nano-scale electronic chemistry,
- Substrates that exhibit self-reconfiguration, fault tolerance, self-repair, and adaptation capabilities,
- Artificial generative and developmental systems,
- Computational intelligence techniques for novel developmental and learning substrates
- Learning through self-organization and emergence
- Julian Francis Miller, Honorary Fellow, University of York, UK
Developmental Approaches to Artificial Neural Network Models
Brains are created by a process of biological development in which neurons and neurites proliferate and change in response to environmental stimuli. Indeed, there is abundant evidence that topological change is a powerful aspect of learning in brains and is responsible for the brain's ability to solve multiple problems.
In contrast artificial neural network models have been idealised as fixed networks of neurons in which learning happens solely through weight adjustment. Such models struggle to solve multiple problems because of catastrophic forgetting whereby learning on one problem is undone by training on another.
This talk concerns a new model of artificial neural networks in which two programs are evolved which control neurons and dendrites. The programs allow dendrites and neurons to die, replicate and change. The programs are represented using Cartesian Genetic Programming. Catastrophic forgetting is avoided and the new approach allows multiple traditional ANNs each solving a different computational problem to be extracted from the underlying artificial brain.
- Alessandro E.P. Villa, Full Professor, University of Lausanne, Switzerland
Evolvable neural networks and attractor dynamics
The sequence of spikes of a neuron carries important information processed by the brain and depends on the subsequent connectivity within the cell assemblies. The topology of such connectivity is the outcome of a neurodevelopmental process associated with genetic as well as context-dependent processes. An association between spatiotemporal patterns of neural discharges and chaotic attractor dynamics was observed in experimental, theoretical and large scale neuronal networks simulations with embedded neuro-developmental features. This talk presents the latest findings of this approach, where the state of the network at any time is represented by the values of control parameters and particular invariant states referred to as attractor states.
Workshop program (Engineer's House Conference Centre, Kronprinsensgate 17, 0251 Oslo)
08:55 - 09:00 Welcome
09:00 - 09:30 Keynote 1: Julian Miller, Developmental Approaches to Artificial Neural Network Models
09:30 - 10:15 Session 1 - 3 x 15 min presentations
10:15 - 10:30 Coffee Break
10:30 - 11:00 Session 2 - 2 x 15 min presentations
11:00 - 11:30 Keynote 2: Alessandro E.P. Villa, Evolvable neural networks and attractor dynamics
11:30 - 12:00 Session 3 - 2 x 15 min presentations (or poster session)
We welcome original contributions describing ongoing projects / completed work (6 pages paper) or preliminary results / novel ideas (3/4 pages brief paper or extended abstract). The instructions for authors and templates can be found at https://icdl-epirob2019.org/submission/. All contributions accepted to the workshop will be invited to submit extended version for a special issue in the journal of Cognitive Neurodynamics, published by Springer. Submissions will be peer reviewed consistent with the journal practices. The papers/abstracts accepted to the workshop will be also published on the workshop webpage.
Submission page: https://ras.papercept.net/conferences/scripts/start.pl#ICDLER19, select -Submit a contribution to ICDL-EPIROB 2019-, then select -Workshop Paper, Submit-, then use the following workshop code: kqxb4
If you have any problems with the submission system, please contact email@example.com
Paper submission: 01 May 2019
Author notification: 01 June 2019
Camera ready version: 01 July 2019
Please feel free to contact us:
Stefano Nichele: firstname.lastname@example.org
Jianhua Zhang: email@example.com
About the organizers