A non-exhaustive overview of research directions I am currently interested in:
-
Representation Geometry
- Geometric structure of representations in artificial and biological neural systems
- Intrinsic dimensionality, subspaces, and higher-order organization
- Dynamics of representations across layers, training, and neural processing
- Cross-system comparison of representational structure
-
Architectural Expressivity and Inductive Bias
- How architecture constrains realizable functions and representations
- Role of depth, gating, recurrence, and composition
- Structural biases induced by architectural design
- Expressivity beyond parameter count
-
Neural Representation and Computation
- Relationship between representation structure and computation
- Population-level organization of neural activity
- Alignment and comparison of representations across subjects, tasks, and models
- Identifying shared computational principles across biological and artificial systems
-
Mechanistic and Causal Analysis
- Mechanisms underlying learned representations and neural computation
- Causal analysis of internal representations and computations
- Memorization, generalization, and task-dependent representation
- Relating internal structure to behavior
-
Foundations of Learning and Representation
- Theoretical limits of representational and computational capacity
- Bridging mathematical theory with empirical phenomena
- Principled measures of representation complexity and expressivity
- Mathematical frameworks applicable across artificial and biological learning systems