Reinhart, René Felix: Reservoir computing with output feedback. 2011
Inhalt
- Introduction
- Associative models, inverse problems and ambiguity
- Bidirectional association
- Forward and inverse models
- Inverse problems and ambiguity resolution
- Flexible selection of models by associative completion
- Reservoir Computing with output feedback
- The Reservoir Computing Paradigm
- Associative Reservoir Computing
- Taxonomy of random projection methods
- Output feedback dynamics and error amplification
- Prospects and challenges
- Programming dynamics of input-driven recurrent neural networks
- Programming dynamics
- State Prediction
- Programming dynamics by predicting states
- Constrained regularization of reservoir networks
- Constrained regularization and State Prediction
- Output feedback stabilization by regularization
- Regularization and stability in reservoir networks with output feedback
- Echo State Networks with output feedback
- Regularization of the read-out layer
- Reservoir regularization
- Regularization stabilizes dynamics and improves performance
- Balancing contributions by distributing activities
- Concluding remarks
- Robust offline learning of associative reservoir networks
- Combined forward and inverse models
- Learning kinematics of a planar robot arm
- Learning kinematics of the Puma robot arm
- Dimensionality reduction and data reconstruction
- Discussion
- Representing and resolving ambiguity with output feedback
- Learning and selecting multiple solutions
- Properties of the dynamical approach to ambiguity
- Forward and inverse kinematics of a planar arm revisited
- Forward and inverse kinematics of the humanoid robot iCub
- Transient- and attractor-based short-term memory
- Concluding remarks
- Movement generation with output feedback dynamics
- Forward and inverse models for movement generation
- Online learning of kinematics for movement generation
- Movement generation with multiple inverse solutions
- Concluding remarks
- Conclusion
- Appendix
- Solving the dual problem
- Reservoir regularization for initially Gaussian distributed weights
- Network initialization and learning parameters
- Kinematics of a planar robot arm with two degrees of freedom
- References
- Related references by the author