Why the Brain Consolidates: Predictive Forgetting for Optimal Generalisation
This work offers a new computational role for offline memory consolidation, explaining phenomena like representational drift and semanticization, which is significant for understanding memory in neuroscience and AI.
The paper proposes that neocortical networks optimize stored representations for generalization by reducing complexity through predictive forgetting, which selectively retains information predicting future outcomes. This process formally improves information-theoretic generalization bounds and is demonstrated in autoencoder, predictive coding, and Transformer models.
Standard accounts of memory consolidation emphasise the stabilisation of stored representations, but struggle to explain representational drift, semanticisation, or the necessity of offline replay. Here we propose that high-capacity neocortical networks optimise stored representations for generalisation by reducing complexity via predictive forgetting, i.e. the selective retention of experienced information that predicts future outcomes or experience. We show that predictive forgetting formally improves information-theoretic generalisation bounds on stored representations. Under high-fidelity encoding constraints, such compression is generally unattainable in a single pass; high-capacity networks therefore benefit from temporally separated, iterative refinement of stored traces without re-accessing sensory input. We demonstrate this capacity dependence with simulations in autoencoder-based neocortical models, biologically plausible predictive coding circuits, and Transformer-based language models, and derive quantitative predictions for consolidation-dependent changes in neural representational geometry. These results identify a computational role for off-line consolidation beyond stabilisation, showing that outcome-conditioned compression optimises the retention-generalisation trade-off.