Phase Attacks Target Vulnerabilities in Complex-Valued Neural Networks

Research into Complex-Valued Neural Networks (CVNNs) identifies Phase Attacks as a highly effective method for reducing model performance. These attacks target phase information specifically, demonstrating that CVNNs are highly susceptible to phase perturbations compared to magnitude-based attacks.