| name | evolutionary-snn-classifier |
| description | Evolutionary feature selection for spiking neural network pattern classifiers using the biologically realistic JASTAP model. Combines evolutionary algorithms with SNN training for simultaneous architecture and feature optimization. |
Evolutionary SNN Classifier with JASTAP
Research methodology from paper "Evolutionary feature selection for spiking neural network pattern classifiers" (2026-04-29).
Core Approach
Applies evolutionary feature selection to the JASTAP (biologically realistic spiking neural network model) for pattern classification tasks.
JASTAP Neural Network Model
- Biologically realistic alternative to standard multi-layer perceptrons
- Incorporates spiking neuron dynamics
- More faithful to biological neural computation than traditional ANNs
Evolutionary Procedure
The paper applies an evolutionary procedure for:
- Simultaneous feature selection - identifying optimal input feature subsets
- Architecture optimization - finding optimal network configurations
- Parameter tuning - optimizing synaptic weights and neuron parameters
Key Benefits
- Reduces input dimensionality automatically
- Finds biologically plausible network architectures
- Avoids manual feature engineering
- Jointly optimizes features and network structure
When to Use
- Classification tasks with high-dimensional input
- Need for biologically plausible neural models
- Scenarios where feature selection is critical
- Applications requiring interpretable feature importance
Workflow
1. Define feature space and JASTAP network architecture
2. Initialize evolutionary population (feature subsets + network configs)
3. Evaluate fitness (classification accuracy + model complexity)
4. Apply selection, crossover, mutation
5. Iterate until convergence
6. Deploy best individual's feature subset + network
Related Skills
spiking-neural-network-analysisbio-neuron-snn-learningmulti-plasticity-snn-training
Paper Reference
- arXiv: 2604.26654
- Authors: Michal Valko, Nuno C. Marques, Marco Castelani
- Date: 2026-04-29
- Categories: cs.NE
