Formal Methods and Machine Learning: Applications, Challenges, and Solutions

Dear Colleagues,

Bridging Machine Learning and Formal Methods for Trustworthy AI.

Machine learning (ML) has rapidly transformed diverse domains such as natural language processing and computer vision, demonstrating unprecedented capabilities in pattern recognition and predictive modeling. Despite these advances, the widespread deployment of ML systems has uncovered significant trust issues in critical safety-sensitive fields like finance, healthcare, and autonomous driving. The complexity, opacity, and unpredictability of models—especially deep neural networks—along with their susceptibility to subtle adversarial attacks, pose serious challenges to certification and safe integration.

While ML offers adaptability and scalability in uncertain environments, formal methods provide rigor, consistency, and guarantees about system behavior. Recently, the intersection of these fields has led to important progress in areas including neural network verification, symbolic learning, probabilistic model checking for ML-based systems, and explainable AI founded on logical constraints.

Nonetheless, formidable challenges remain. Formal verification techniques struggle to scale efficiently to deep and complex architectures, and it remains difficult to formally express intricate learning objectives within formal languages. This Special Issue invites contributions that investigate the challenges, innovations, and applications at this intersection.

The topics of interest encompass the following:

• Using runtime assurance, model checking, and formal verification to ensure ML models fulfill stringent safety, fairness, and robustness standards.
• Exploring how machine learning can automate and scale formal analysis processes.
• Developing end-to-end system assurance methodologies.
• Defining meaningful formal properties tailored to learning systems.
• Scaling verification frameworks to handle large-scale ML models.

By merging formal reasoning's rigor with ML's data-driven strengths, this Special Issue seeks to establish a roadmap toward responsible, reliable, and provably safe intelligent systems, an essential step for building trustworthy AI.

Topics of interest:

• The formal verification and validation of machine learning models.
• The formal verification of deep neural networks in safety-critical domains.
• Formal methods for reinforcement learning safety.
• The certification and assurance of ML components.
• The integration of formal methods into AI model development pipelines.
• Symbolic and logic-based reasoning for machine learning systems.
• Explainability, interpretability, and accountability through formal analysis.
• Formal specification and analysis for autonomous and intelligent systems.
• Formal specification learning and requirement mining.
• Model checking and runtime assurance for adaptive AI systems.
• Hybrid approaches: Combining learning and formal reasoning.
• Scalability and efficiency in ML verification.
• Challenges and future directions in trustworthy and certifiable AI.

Prof. Dr. Kamel Barkaoui
Prof. Dr. Zhiwu Li
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Information is an international peer-reviewed open access monthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 1800 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

• formal verification
• machine learning
• safety
• robustness
• adversarial robustness
• certified intelligence
• explainable artificial intelligence
• neural network verification
• trustworthy AI
• model checking
• logic-guided learning
• safety-critical AI applications