Rethinking Reservoir Pruning: A Dynamical Perspective for Echo State Networks
Quick summary
arXiv:2608.04593v1 Announce Type: cross Abstract: Echo State Networks (ESNs) offer an efficient framework for temporal prediction, but their randomly initialized reservoirs are often over-parameterized and dynamically redundant. Existing pruning methods largely rely on static connectivity or activation statistics, which may overlook neurons that shape input-driven state transitions. We propose Dynamical Mode Pruning (DMP), a reservoir pruning method that ranks neurons by their contribution to dominant transition modes obtained from a trajectory-averaged Jacobian Gramian. DMP removes low-impact
Key takeaways
- arXiv:2608.04593v1 Announce Type: cross Abstract: Echo State Networks (ESNs) offer an efficient framework for temporal prediction, but their randomly initialized reservoirs are often over-parameterized and dynamically redundant.
- Existing pruning methods largely rely on static connectivity or activation statistics, which may overlook neurons that shape input-driven state transitions.
- We propose Dynamical Mode Pruning (DMP), a reservoir pruning method that ranks neurons by their contribution to dominant transition modes obtained from a trajectory-averaged Jacobian Gramian.
Why it matters
The importance of “Rethinking Reservoir Pruning: A Dynamical Perspective for Echo State Networks” will be measured by what changes in practice. User behavior, access conditions, verifiable performance and responsible-use outcomes are the signals worth following.
Member comments