Multiobjective optimization of echo state networks for multiple motor pattern learning

摘要: Echo State Networks are a special class of recurrent neural networks, that well-suited for attractorbased learning motor patterns. Using structural multiobjective optimization, the trade-off between network size and accuracy can be identified. This allows to choose feasible model capacity follow-up full-weight optimization. It is shown produce small efficient capable storing multiple patterns in single net. Especially smaller networks interpolate learned using bifurcation inputs.