How can models learn continuously without catastrophic forgetting? Exploring memory-augmented architectures, rehearsal strategies, and dynamic network expansion for lifelong learning systems.

Beyond correlation: building models that understand cause and effect. Investigating causal inference in neural networks, interventions, and counterfactual reasoning for more interpretable AI.

Rethinking backpropagation: exploring biological plausibility, energy-based models, and novel optimization approaches. Can we build learning systems that are more efficient and robust?