Warm-up decrements are ubiquitous and substantial in naturalistic learning: the two-timescale model M2 improves out-of-sample prediction across the board over the baseline model M1¶

Claim · All claims
Evidence · 1 study · associational r2 · q2
Currency · newest study 2022 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? M2 provides a substantial, across-the-board improvement over M1 in out-of-sample RMSE, revealing the ubiquity of warm-up decrements. → Aakriti Kumar 2022

Evidence¶

Aakriti Kumar 2022¶

Aakriti Kumar, Aaron S. Benjamin, Andrew Heathcote and Mark Steyvers. (2022). Comparing models of learning and relearning in large-scale cognitive training data sets. npj Science of Learning 7:24. https://doi.org/10.1038/s41539-022-00142-x

q2 · i? · associational · r2

Predictive model comparison on held-out Lumosity data (Table 1): for Lost in Migration, M2 RMSE 3.56 vs M1 3.61 on all data; for Ebb and Flow, 4.54 vs 4.56. Neither simple model was competitive overall.

"Note first that both the single-parameter learning model ( M1) and the traditional two-timescale model ( M2) are not competitive in accounting for learning in this data set. Yet M2 provides a substantial, across-the-board improvement over M1. This outcome reveals the ubiquity and substantial nature of warm-up decrements in naturalistic learning."

Discussion¶

Related Claims¶

• The interactive two-timescale model (M4) provides the most general account of naturalistic learning, though the additive-forgetting model (M3) fits better at six of eight interval subsets — related