The Memory of Attractors in Recurrent Neuronal Networks and its Application to Human-Machine Interaction


  • Ron Goerz University of Vienna


This project is about the practical implementation of a recurrent neuronal network with the goal of using oscillating attractors as a model for memory in an embedded system. Through the neuronal simulation, it is shown that the properties of dynamic systems can provide a model for memory.

Aram et al. [1] suggest to interpret the oscillating output-states (oscillating attractors) of recurrent neuronal networks as a storage for information. By changing one parameter of the system, it is possible to access different oscillating attractors which are interpreted as different stored information. This parameter represents the level of dopamine in the brain, whereby dopamine is associated with attention and focusing.

The suggested model of Aram et al. [1] is embedded in a controlling system of a robotic arm to prove its applicability in the context of a simulated environment and the plausibility of the memory system.

To implement the system, a replication of the model of [1] is done in a first step. In a second step, the outputs of the memory system are connected to the virtual actuators of a digital simulation of a robotic arm. In the following, the system is trained and the behaviour observed.

On the one hand, it can be demonstrated that recurrent neuronal networks can store information and provide a model for memory in biological organisms, which consist of recurrent neuronal networks. On the other hand, a controlling system is suggested, which provides a neurobiological inspired approach to operate a robotic arm in a simulation.


[1] Z. Aram, S. Jafari, J. Ma, J. C. Sprott, S. Zendehrouh, and V.-T. Pham, “Using chaotic artificial neural networks to model memory in the brain”, Communications in Nonlinear Science and Numerical Simulation, vol. 44, pp. 449–459, 2017, doi: /10.1016/j.cnsns.2016.08.02 5.