Working definition
A spiking neural network is a directed dynamical system whose units integrate inputs, evolve internal state, emit discrete spikes under a defined rule, and influence other units through weighted, delayed connections. Models differ substantially in neuron equations, coding, training, and hardware mapping; use of spikes alone does not make a model biologically realistic or computationally superior.
Mechanism
- Integrate synaptic inputs into neuron state.
- Emit a spike when a specified condition is met.
- Propagate weighted and delayed events through a graph.
Measurements
- Accuracy or task loss
- Spike count and temporal latency
- Memory and operation count
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
- Training remains task- and model-dependent.
- Sparse spikes do not guarantee low end-to-end energy.
Mathematical connection
Formal structure without substrate erasure
State, threshold, and reset dynamics
Represent neuron and network evolution with declared state equations and event conditions.
Inputs
- Neuron equations
- Connection graph
- Initial state
Outputs
- State trajectory
- Spike events
- Stability conditions
Limit: A dynamical model can reproduce selected signals without reproducing biological mechanism, cognition, or subjective experience.
Technical and governance sources
- [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.
- [2]Taking Neuromorphic Computing to the Next Level with Loihi 2 · Intel Labs
Establishes: An official description of the Loihi 2 research chip, its programmable neuron models, event-based communication, on-chip learning support, and the Lava software framework used to construct neuromorphic applications.
Boundary: This is a vendor technical brief about a research platform. Performance and efficiency results remain workload-, configuration-, measurement-boundary-, and comparison-dependent and do not establish equivalence to biological intelligence.