By name: quantum-linear-reservoir-bottleneck description: "Analysis of hidden bottleneck in classical and quantum linear reservoir computing. Identifies fundamental information processing capacity limits when reservoir features and readout are both linear. Use when: reservoir computing design, quantum reservoir computing, linear system capacity analysis, information processing capacity bounds, echo state networks, quantum machine learning architecture design." license: Complete terms in LICENSE.txt metadata: arxiv_id: "2605.29071" published: "2026-05-29" tags: [quantum, reservoir-computing, linear-systems, information-capacity, machine-learning]
Hidden Bottleneck in Linear Reservoir Computing
Core Methodology
Identifies a fundamental bottleneck in linear reservoir computers (both classical and quantum): when measured features evolve linearly in the reservoir and the output is formed by linear readout, the information processing capacity is severely limited regardless of reservoir size.
Key Insights
- Linearity Bottleneck: Linear reservoir + linear readout = fundamentally limited expressivity, even with infinite reservoir size
- Quantum Does Not Help: Quantum linear reservoirs suffer from the same bottleneck—quantumness alone does not break the linearity limit
- Capacity Bound: Derives explicit upper bounds on information processing capacity for linear reservoir systems
- Design Implication: Nonlinear feature extraction is essential—either in the reservoir dynamics or the measurement/readout
Mathematical Result
For a linear reservoir with state evolution x(t+1) = Ax(t) + Bu(t) and linear readout y(t) = Cx(t):
- The effective memory depth is bounded by the rank of the observability matrix
- Adding more reservoir dimensions does not increase capacity beyond this bound
- The bottleneck is structural, not a matter of training or optimization
When to Use
- Designing reservoir computing systems (classical or quantum)
- Analyzing why a linear reservoir underperforms
- Deciding whether to add nonlinear features to a reservoir
- Understanding fundamental limits of linear dynamical systems for ML
Practical Recommendations
- Add nonlinearity in measurement: Use nonlinear observables/measurement operators
- Add nonlinearity in dynamics: Introduce nonlinear state transitions
- Kernel trick: Map linear features to nonlinear feature space before readout
- For quantum reservoirs: Use nonlinear measurement bases, not just computational basis
Related Approaches
- Echo State Networks (ESNs) with nonlinear activation
- Liquid State Machines
- Quantum reservoir computing with nonlinear readout
- Kernel methods for dynamical systems