Search

Abstract: Reservoir computing is a framework for computation like a recurrent neural network that allows for the black box modeling of dynamical systems. In contrast to other recurrent neural network approaches, reservoir computing does not train the input and internal weights of the network, only the readout is trained. However it is necessary to adjust parameters to create a “good” reservoir for a given application. In this study we introduce a method, called RCDESIGN (reservoir computing and design training). RCDESIGN combines an evolutionary algorithm with reservoir computing and simultaneously looks for the best values of parameters, topology and weight matrices without rescaling the reservoir matrix by the spectral radius. The idea of adjust the spectral radius within the unit circle in the complex plane comes from the linear system theory. However, this argument does not necessarily apply to nonlinear systems, which is the case of reservoir computing. The results obtained with the proposed method are compared with results obtained by a genetic algorithm search for global parameters generation of reservoir computing. Four time series were used to validate RCDESIGN. [Copyright &y& Elsevier]

Biologically inspired controllers demonstrate great success in several applications, mainly in situations that present disturbances and uncertainties in the dynamics of the system. In recent times, several works have appeared in the area of emotional learning which occurs in the human brain, thus allowing to the emergence of new theories and applications in control engineering. In control engineering, it is possible to highlight the BELBIC (Brain Emotional Learning-Based Intelligent Controller). However, the design and commissioning of this type of controller still represents a major challenge for researchers, since it is necessary to determine some characteristic signals to this system (stimuli), which can vary from application to application. This work presents a methodology for the construction of architectures for BELBIC stimulus signals, using as a basis the DRL (Deep Reinforcement Learning) techniques. The DRL allows extracting characteristic patterns from the dynamics of systems which, perhaps, may have high dimensionality and possibly nonlinear dynamics, as is the case of most problems involving real-world dynamic systems. The resulting controller model is validated by applying an inverted pendulum dynamic system in order to demonstrate a new approach to the architectures of the BELBIC that allows to achieve a greater generalization in its application, as well as providing a viable alternative to the traditional models in use. [ABSTRACT FROM AUTHOR]