Tight neural network verification via semidefinite relaxations and linear reformulations
We present a novel semidefinite programming (SDP) relaxation that enables tight and
efficient verification of neural networks. The tightness is achieved by combining SDP …
efficient verification of neural networks. The tightness is achieved by combining SDP …
Repairing misclassifications in neural networks using limited data
We present a novel and computationally efficient method for repairing a feed-forward neural
network with respect to a finite set of inputs that are misclassified. The method assumes no …
network with respect to a finite set of inputs that are misclassified. The method assumes no …
Robust explanations for human-neural multi-agent systems with formal verification
F Leofante, A Lomuscio - European Conference on Multi-Agent Systems, 2023 - Springer
The quality of explanations in human-agent interactions is fundamental to the development
of trustworthy AI systems. In this paper we study the problem of generating robust contrastive …
of trustworthy AI systems. In this paper we study the problem of generating robust contrastive …
Leveraging satisfiability modulo theory solvers for verification of neural networks in predictive maintenance applications
Interest in machine learning and neural networks has increased significantly in recent years.
However, their applications are limited in safety-critical domains due to the lack of formal …
However, their applications are limited in safety-critical domains due to the lack of formal …
Verification of semantic key point detection for aircraft pose estimation
Abstract We analyse Semantic Segmentation Neural Networks running on an autonomous
aircraft to estimate its 6DOF pose during landing. We show that automated reasoning …
aircraft to estimate its 6DOF pose during landing. We show that automated reasoning …
Formal and Practical Elements for the Certification of Machine Learning Systems
Over the past decade, machine learning has demonstrated impressive results, often
surpassing human capabilities in sensing tasks relevant to autonomous flight. Unlike …
surpassing human capabilities in sensing tasks relevant to autonomous flight. Unlike …
Verification-friendly networks: the case for parametric relus
F Leofante, P Henriksen… - 2023 International Joint …, 2023 - ieeexplore.ieee.org
It has increasingly been recognised that verification can contribute to the validation and
debugging of neural networks before deployment, particularly in safety-critical areas. While …
debugging of neural networks before deployment, particularly in safety-critical areas. While …
NeVer2: learning and verification of neural networks
NeVer2 is an open-source, cross-platform tool aimed at designing, training, and verifying
neural networks. It seamlessly integrates popular learning libraries with our verification …
neural networks. It seamlessly integrates popular learning libraries with our verification …
Formal Verification of Neural Networks: A “Step Zero” Approach for Vehicle Detection
This paper delves into the verification of Convolutional Neural Networks for the crucial task
of identifying vehicles in automotive images. Given the complexity and verifiability …
of identifying vehicles in automotive images. Given the complexity and verifiability …
Verifying autoencoders for anomaly detection in predictive maintenance
In recent years, the application of artificial intelligence and machine learning techniques has
gained significant traction in addressing various challenges across industries. Among these …
gained significant traction in addressing various challenges across industries. Among these …