Information coding strategies in biological reservoir computing for multiclass time-series classification

Latency coding outperformed pulse amplitude modulation for multiclass time-series classification in primary rat cortical neuron reservoir computing, improving 10-class spoken digit accuracy by 10%.

• Why it matters: It remains unclear how to design electrical inputs and extracellular readouts so biological neural networks can support complex information processing under different task demands.
• What they did: In primary rat cortical neurons on a high-density microelectrode array, the study compared pulse amplitude modulation versus latency coding and tested spike sorting’s effect on spike detection and class separability.
• The result: Spike sorting improved classification performance by enhancing class separability while largely preserving the low-dimensional structure of population activity, supporting task-dependent coding design.