Finding good Echo State Networks to control an underwater robot using evolutionary computations

Training Recurrent Networks by Evolino

“…Evolino evolves weights to the nonlinear, hidden nodes while computing optimal linear mappings from hidden state to output, using methods such as pseudo-inverse-based linear regression (Penrose, 1955) or support vector machines (Vapnik, 1995), depending on the notion of optimality employed. This generalizes methods such as those of Maillard (Maillard & Gueriot, 1997) and Ishii et al (Ishii, van der Zant, Bečanović, & Plöger, 2004;van der Zant, Bečanović, Ishii, Kobialka, & Plöger, 2004) that evolve radial basis functions and ESNs, respectively. Applied to the LSTM architecture, Evolino can solve tasks that ESNs (Jaeger, 2004a) cannot and achieves higher accuracy in certain continuous function generation tasks than conventional gradient descent RNNs, including gradient-based LSTM (G-LSTM).…”

Reservoir Computing: Size and Connectivity Optimization using the "Worm Algorithm"

“…Both short time and low number of computations are met in this algorithm; unlike the methods used in [7] and [8] even thought these methods were used to optimize the Spectral radius as well. It would be very useful if the designers consider using it to find the lowest RS and CP that can perform the required task before the hardware implementation, instead of making their decisions on experts' estimations.…”

“…The most appropriate genetic algorithm was used in [7] and [8] with a similar target. In that work Evolutional algorithm (EA) [9] and Evolutional Strategy (ES) [10] were used to find an ESN to control underwater robot motion in [11].…”

Supervised learning of internal models for autonomous goal-oriented robot navigation using Reservoir Computing

“…In these context, the current work is the first to integrate localization and navigation using the Reservoir Computing paradigm. Several recent works have been using RC in robotics: in [16] for mobile robot modeling and control, in [10] for movement generation, in [17] for motor control, and in [22] for underwater robot control.…”