Silicon and physical substrates

Physical reservoir computing

Use a physical system’s nonlinear transient dynamics as a feature transformation for a trained readout.

emerging deviceexperimental platform

Working definition

Physical reservoir computing drives a dynamical substrate with time-varying input, samples its high-dimensional transient response, and trains a comparatively simple readout. The substrate may be photonic, mechanical, magnetic, electronic, chemical, or biological. Evaluation requires stability, memory capacity, controllability, readout cost, reset behavior, and fair digital baselines.

Mechanism

  • Inject a time-varying input into a nonlinear substrate.
  • Measure an expanded transient state.
  • Train and validate a readout on held-out sequences.

Measurements

  • Task loss
  • Memory capacity
  • Drift, reset time, and readout cost

Reproducibility controls

  • Version hardware, software, firmware, and analysis code.
  • Declare dataset, preprocessing, random seeds, and measurement boundary.
  • Report repeated runs, variation, exclusions, and failed trials.

Limits and failure modes

  • Physical complexity is not useful computation by itself.
  • Readout and instrumentation may dominate.

Mathematical connection

Formal structure without substrate erasure

computational modelDynamical systems

Transient substrate response

Characterize nonlinear response, fading memory, stability, and separability for a declared input regime.

Inputs

  • Input sequence
  • Measured substrate state
  • Readout protocol

Outputs

  • State-space features
  • Memory estimate
  • Stability region

Limit: Rich dynamics are not useful computation until a held-out task and complete readout cost demonstrate value.

Technical and governance sources

  1. [1]NeuroBench: Advancing Neuromorphic Computing Through Collaborative, Fair and Representative Benchmarking · National Institute of Standards and Technology

    Establishes: A community framework separating algorithm and system tracks and defining task, correctness, efficiency, and reporting procedures intended to make neuromorphic results more comparable and reproducible.

    Boundary: A benchmark ranks submitted systems on declared tasks and metrics. It does not prove general intelligence, biological equivalence, safety, usefulness outside the benchmark, or superiority under unreported host and data costs.

Related concepts