Statements (46)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
Computational paradigm
|
| gptkbp:advantage |
Efficient training
Handles temporal data |
| gptkbp:application |
gptkb:robot
gptkb:signal_processing Control systems Speech recognition Time series prediction |
| gptkbp:component |
gptkb:reservoir
Input layer Readout layer |
| gptkbp:feature |
gptkb:Nonlinear_dynamics
Fixed recurrent network Low training complexity Trains only output weights |
| gptkbp:field |
gptkb:artificial_intelligence
Machine learning Neural networks |
| https://www.w3.org/2000/01/rdf-schema#label |
Reservoir Computing
|
| gptkbp:inspiredBy |
Biological neural networks
|
| gptkbp:introducedIn |
Early 2000s
|
| gptkbp:limitation |
Performance depends on reservoir properties
Reservoir design is nontrivial |
| gptkbp:notableContributor |
gptkb:Herbert_Jaeger
gptkb:Wolfgang_Maass |
| gptkbp:notableModel |
gptkb:Echo_State_Network
Liquid State Machine |
| gptkbp:notablePublication |
The 'Echo State' Approach to Analysing and Training Recurrent Neural Networks (Jaeger, 2001)
A Model for Real-Time Computation in Generic Neural Microcircuits (Maass et al., 2002) |
| gptkbp:relatedConcept |
Spiking neural networks
Backpropagation through time Computational neuroscience Feedforward neural networks Nonlinear system identification Temporal processing |
| gptkbp:relatedTo |
gptkb:Echo_State_Networks
gptkb:Recurrent_neural_networks Liquid State Machines Memristor-based reservoir computing Optical reservoir computing Physical reservoir computing |
| gptkbp:uses |
Nonlinear activation functions
Randomly connected network Sparse connectivity |
| gptkbp:bfsParent |
gptkb:Herbert_Jaeger
|
| gptkbp:bfsLayer |
7
|