Leveraged Learning: entropy cleared per bit received
Abstract
An expository essay containing some new results. A learner holds a prior belief over Boolean maps that answer a finite set of Q questions, and receives answers one by one. Each answer carries surprisal and can clear predictive uncertainty about both the question asked and questions still unasked. We quantify this effect by the leverage: table entropy cleared per bit of surprisal received. Taking expectations over the truth prior and the question order, define the aggregate leverage as the ratio of expected uncertainty cleared to expected surprisal received. It equals one for independent answers and can exceed one for correlated answers. At finite size, this ratio is determined exactly by a single sequence: the mean entropy Gℓ of the answers to ℓ questions. As the number of input bits grows at fixed asked fraction t=ℓ/Q, a limiting increment profile γ(t) determines the macroscopic learning curve. With η0 the limiting initial entropy per question, the leverage becomes L(t)= (η0−(1−t)γ(t))/∫0tγ(x),dx. For exchangeable priors, de Finetti's representation gives a constant bulk profile γ(t): deduction is confined to a boundary layer at t=0, and the leverage is forced to a hyperbolic form, as surprisal grows linearly. By contrast, we construct a simplicity prior with nontrivial bulk learning by grading Boolean maps by their polynomial degree over 𝔽2 and allocating weight across degree classes through a CDF F. Reed-Muller capacity then yields γ(t)=1−F(t). This realizes any nonincreasing profile taking values in [0,1], together with the corresponding macroscopic leverage curve.
Related
- Entropy-Regularized Mean-Field Games with Average-Reward Criterion: Span-Seminorm Contractions and Model-Free Learning
- Learning from Totally Unpaired Data via Minimum Conditional Entropy Coupling
- Entropy-Regularized Reinforcement Learning for the Regime-Switching Consumption-Investment Problem
- Human-Centered Learning Mechanics: Observable Entropy Control under Protected-Risk Constraints
- Conditional Flow Models for Offline Maximum Entropy Reinforcement Learning
BibTeX
@misc{chernowitz2026leveraged,
title = {{Leveraged Learning: entropy cleared per bit received}},
author = {Daniel Chernowitz},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.08054},
eprint = {2609.08054},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.08054},
}
RIS
TY - UNPB
TI - Leveraged Learning: entropy cleared per bit received
AU - Daniel Chernowitz
PY - 2026
DA - 2026/09/30
PB - arXiv
DO - 10.48550/arxiv.2609.08054
UR - https://arxiv.org/abs/2609.08054
AB - An expository essay containing some new results. A learner holds a prior belief over Boolean maps that answer a finite set of Q questions, and receives answers one by one. Each answer carries surprisal and can clear predictive uncertainty about both the question asked and questions still unasked. We quantify this effect by the leverage: table entropy cleared per bit of surprisal received. Taking expectations over the truth prior and the question order, define the aggregate leverage as the ratio of expected uncertainty cleared to expected surprisal received. It equals one for independent answers and can exceed one for correlated answers. At finite size, this ratio is determined exactly by a single sequence: the mean entropy G_ℓ of the answers to ℓ questions. As the number of input bits grows at fixed asked fraction t=ℓ/Q, a limiting increment profile γ(t) determines the macroscopic learning curve. With η_0 the limiting initial entropy per question, the leverage becomes L(t)= (η_0−(1−t)γ(t))/∫_0^tγ(x),dx. For exchangeable priors, de Finetti's representation gives a constant bulk profile γ(t): deduction is confined to a boundary layer at t=0, and the leverage is forced to a hyperbolic form, as surprisal grows linearly. By contrast, we construct a simplicity prior with nontrivial bulk learning by grading Boolean maps by their polynomial degree over 𝔽_2 and allocating weight across degree classes through a CDF F. Reed-Muller capacity then yields γ(t)=1−F(t). This realizes any nonincreasing profile taking values in [0,1], together with the corresponding macroscopic leverage curve.
ER -