Evolving explainable neural architectures: Dynamic behavioral-novelty weighting in multi-objective neuroevolution for multimodal spam detection

Key Points

• The algorithm achieved a 25% increase in spam detection accuracy compared to traditional methods, showcasing key improvements in performance.
• Assessment using multi-objective neuroevolution facilitated the dynamic adjustment of behavioral-novelty weighting to enhance detection capabilities.
• Through innovative explainable neural architectures, the framework adapts to various modalities, tackling diverse spam types effectively.
• Findings indicate that optimizing behavioral-novelty weighting may enable more robust spam filtering across different data types.