Research
120d agoCausal Analysis Explains Binary Spiking Neural Networks
Researchers represented Binary Spiking Neural Networks as binary causal models to decode their internal behavior. By applying SAT and SMT solvers to the MNIST dataset, the team extracted abductive explanations for specific classifications. This logic-based approach offers a more transparent alternative to SHAP.
The Signal
It provides a concrete path for interpreting non-traditional neural architectures.