PhD Position in Trustable Machine Learning: Analysis and Verification of Soft Automata

Inria Rennes invites applications for a three-year doctoral position focusing on trustable machine learning and the analysis of soft automata. The successful candidate will join the collaborative SAIF project to investigate the theoretical and experimental properties of neural network-based dynamic models, bridging statistics, optimization, and formal verification. The work involves exploring diverse architectures ranging from recurrent neural networks to modern state space models and transformers. Candidates are expected to conduct rigorous research, prototype novel algorithms, publish in top-tier venues, and actively participate in project meetings and scientific training.

• Conduct comprehensive literature reviews on machine learning verification and automata learning.
• Perform theoretical research on the properties, expressivity, and robustness of soft automata.
• Design and prototype original algorithms using standard machine learning libraries.
• Analyze simulation results and evaluate model sensitivity to noise and adversarial attacks.
• Write research papers for submission to international conferences and journals.
• Present research findings at project meetings, team discussions, and academic conferences.
• Complete over 100 hours of mandatory scientific training over the course of the three-year thesis.
• Draft the final doctoral dissertation and defend the thesis.