biological-plausibility-assessment-snn

name	biological-plausibility-assessment-snn
description	Automated framework for assessing biological plausibility of spiking neuron models using Izhikevich firing pattern classification. Use when evaluating spiking neuron models, comparing neuromorphic designs, or quantifying how well artificial neurons replicate biological behavior.
metadata	{"arxiv_id":"2606.17853","published":"2026-06-16","authors":"","tags":["spiking-neuron","biological-plausibility","izhikevich","neuromorphic","optimization"]}

Biological Plausibility Assessment Framework

Core Idea

Automated black-box assessment of spiking neuron models by evaluating their ability to replicate canonical Izhikevich firing patterns through parameter optimization, without requiring analytical models.

Izhikevich Firing Patterns (20 canonical patterns)

The framework tests against these biological firing behaviors:

• Tonic spiking, Phasic spiking, Tonic bursting, Phasic bursting
• Delayed spiking, Transient spiking, Spike frequency adaptation
• Class 1, Class 2, Spike resonance, Accommodation
• Subthreshold oscillation, Bistability, Depolarizing block
• Threshold variability, Rebound spike, Rebound burst
• Inhibition-induced spiking, Inhibition-induced bursting

Assessment Workflow

1. Define target patterns: Select which Izhikevich patterns to test
2. Encode as objective functions: Each pattern becomes an optimization target
3. Black-box optimization: Optimize neuron parameters to match each pattern
4. Scoring: Measure how well the model reproduces each pattern
5. Aggregate plausibility score: Combine individual pattern scores

Implementation (Python/PyTorch/Norse)

# Pseudocode for the assessment framework
def assess_plausibility(neuron_model, target_patterns):
    scores = {}
    for pattern in target_patterns:
        # Encode pattern as objective function
        objective = encode_pattern(pattern)
        # Optimize model parameters
        optimal_params = optimize(neuron_model, objective)
        # Score the match
        scores[pattern] = evaluate_match(neuron_model, optimal_params, pattern)
    return aggregate_score(scores)

Applications

• Neuron model selection: Choose most biologically plausible model for a task
• Neuromorphic hardware validation: Verify hardware neuron dynamics match biology
• Research benchmarking: Compare new neuron models against established baselines
• Architecture design: Guide design of neuromorphic systems

Pitfalls

• Pattern coverage: Limited to Izhikevich's 20 patterns — may miss other biological behaviors
• Optimization sensitivity: Results depend on optimizer choice and hyperparameters
• Black-box limitation: Does not explain WHY a model fails, only that it does
• Computational cost: Full assessment requires optimizing for each pattern separately
• Parameter bounds: Results sensitive to parameter search ranges

Activation Keywords

• biological plausibility
• neuron model assessment
• izhikevich patterns
• spiking neuron evaluation
• neuromorphic validation
• firing pattern matching
• neuron model benchmark