Nature has, through evolution, developed information processing in living organisms that we aim to use to increase our ability to computationally handle the increasing amounts of data that are being generated. Therein lies the importance of knowledge of natural computation processes modeled in scientific frameworks and with analytical tools.
Computation can be understood as physical processes in nature. Natural computation can be used to explain emergent phenomena by complexification of information through computational processes at different levels of organization. We propose a synthetic framework in which information represents the structure and computation processes its changes (dynamics). The consequence of the new framework is that physical objects and processes can be modeled, interpreted and predicted by info-computational methods. The new framework will be used as a tool for studying cognitive systems such as living organisms on different levels of complexity.
MORCOM@COGS develops a conceptual framework in which the computation process is generalized from abstract symbol manipulation of the Turing machine type to the information processes in physical systems. The project studies how information is created and structured on different levels or scales and how it changes through natural processes in cognitive systems.
Consequences of the shift in modeling of computation toward cognitive computing are examined by comparing existing models with the new, morphological computations applied to various classes of cognitive systems in nature. Complex systems in nature have already inspired a number of methods for information processing - including artificial neural networks, genetic algorithms and genetic programming, and development continues.
The project aims to:
1) Further develop the synthetic framework of natural info-computation. There are already enough elements of a theory based on natural computation processes that is much more appropriate for modeling of complex cognitive systems than those we have today. We compare the new approach with the existing Turing machine model - open versus closed systems, generative versus predetermined sequence, parallel versus sequential processes, and so on. We investigate the consequences of the generalization of the concept of computation with application to morphological processes in which the morphology of a cognitive system determines the dynamics of its information structure.
2) The implementation project is about applying morphologically based models to various cognitive systems. In morphologically based models sensors and motors are given by the morphology of the genotype, while the middle layer develops in an ongoing process as phenotype. An organism consists here of an information network of networks developed according to a few basic principles, sensors and motors, morphology-based Hebbian and Bayesian learning, perception and anticipation, and so on. This leads to a morphological system that constantly evolves in interaction with the environment.
In the project we will focus on the following three processes:
• Sensory-to-motor transformation of data through self-organization of information as morphological computation
• Learning and memory by morphology
• Decision making, fast and slow (Kahneman dual-system process theory) by signal vs. symbol processing, based on morphological computations.
The above applies to different classes of cognitive agents - from the simplest single-celled organisms and their networks to the most complex agents such as humans.
The project is organized into the following work packages:
1. The info-computational framework development for cognitive systems
2. Morphological computation as a model of cognition
3. Implementation of morphological computation to different types of cognitive systems
4. Synthesis and dissemination
The results of the work packages will enable us to improve the basic framework as well as our understanding of the different types of cognitive modeling.
Docent vid Applied Information Technology (Chalmers)
Funding Chalmers participation during 2016–2020
Areas of Advance