PhD thesis on the robustness of dynamical systems for learning in games (M/F)

Location: POLARIS team (https://team.inria.fr/polaris/), a joint research team between CNRS, Inria and Université Grenoble Alpes, part of the LIG Laboratory, a mixed research unit (UMR5217) of the CNRS (450 people). The team is located on the Saint Martin d'Hères campus, accessible by tram from Grenoble.

Other benefits:
- 45 days annual leave
- Partial coverage of complementary health insurance costs.
- Subsidized catering available on campus
- Partial reimbursement of public transport costs

From automated hospital admission systems powered by machine learning (ML), to flexible chatbots capable of fluent conversations and self-driving cars, the wildfire spread of artificial intelligence (AI) has brought to the forefront a crucial question with far-reaching ramifications for the society at large: Can ML systems and models be relied upon to provide trustworthy output in high-stakes, mission-critical environments?

These questions invariably revolve around the notion of robustness, an operational desideratum that has eluded the field since its nascent stages. One of the main reasons for this is the fact that ML models and systems are typically data-hungry and highly sensitive to their training input, so they tend to be brittle, narrow-scoped, and unable to adapt to situations that go beyond their training envelope. On that account, robustness cannot be achieved by blindly throwing more data and computing power to larger and larger models with exponentially growing energy requirements (and a commensurate carbon footprint to boot). Instead, this thesis proposal intends to focus on the core theoretical and methodological foundations of robustness required for current and emerging AI systems.

In more detail, to address the challenges that arise when ML models and algorithms are deployed and interact with each other in real-life environments, we plan to develop the required theoretical and technical tools for AI systems that are able to (a) adapt “on the fly” to non-stationary environments; and (b) gracefully interpolate from best- to worst-case guarantees. In particular, this thesis intends to focus on the replicator dynamics, a particularly fruitful model of multi-agent learning in games, the overarching objective being to obtain a complete description of its robustness to noise and uncertainty. Familiarity and experience with the replicator dynamics will be a prerequisite for this PhD.