<ul><li>Recent works have focused on studying adversarial robustness of neural networks for resource-constrained embedded systems.</li><li>A new neural network conversion algorithm has been introduced to create sparse and adversarially robust spiking neural networks (SNNs).</li><li>The algorithm leverages sparse connectivity and weights from a robustly pretrained artificial neural network (ANN).</li><li>The approach combines energy efficiency of SNNs with the novel conversion algorithm, resulting in improved performance and robustness against adversarial threats.</li></ul>

Adversarially Robust Spiking Neural Networks with Sparse Connectivity

Discover more