VNN 2020 - Verification of Neural Networks

Explore cutting-edge research in neural network verification through this comprehensive workshop from the Conference on Computer-Aided Verification. Delve into advanced techniques for ensuring the reliability and safety of AI systems across various applications, from cyber-physical systems to adversarial environments. Learn about scalable verification abstractions for AI-controlled systems and discover methods for end-to-end robustness in sensing-reasoning pipelines when facing adversarial attacks. Examine innovative approaches including image approximation techniques for feed-forward neural networks, execution-guided overapproximation (EGO) methods, and property inference using linear interpolants in ReLU networks. Investigate L1-norm channel pruning strategies specifically designed for convolutional neural network verification, and participate in live discussions with leading researchers in the field. Gain insights into the latest developments in formal verification methods, safety guarantees, and robustness analysis that are essential for deploying neural networks in critical applications where reliability and trustworthiness are paramount.