A Unifying Framework From Neural Superposition to Sparse Interpretable Codes

A new Nature Machine Intelligence paper proposes a unifying framework linking neural-network superposition to sparse, interpretable codes.