← 研踪 PaperTrailarXiv:2611.24881 · v1
Monoculture Robust Learning
单一栽培鲁棒学习
首次提交 2026-11-19 02:27
版本更新 2026-11-19 02:27
雷达观测 2026-11-19 15:04
摘要
We introduce a notion of monoculture robustness in Probably Approximately Correct (PAC) learning, as a way to mitigate the problem of algorithmic monoculture (Kleinberg and Raghavan, 2021): the tendency of models sharing the same components (e.g., training data) to fail on the same inputs. Specifically, we ask whether a learner with access to a single dataset can output individually accurate hypotheses whose joint failure probability is comparable to training on independent datasets. We say that such a learning procedure achieves monoculture robustness, and we quantify this by bounding the conjunctive error that all hypotheses misclassify points in an individual sense. We design methods for achieving this type of guarantee by drawing a surprising connection between replicability (Impagliazzo et al., 2022) and monoculture robustness. Specifically, we show that if the base learner is -replicable, then running it independently times on the same sample yields individual monoculture error that shrinks exponentially with , controlled by the single-run error of the base learner at the point and . We complement these positive results with a lower bound: for individual monoculture robustness, we prove a worst-case lower bound showing that, for learners, driving the individual conjunctive error down to can require a sample-size multiplier of .
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-19 02:27 · 最新观测版本
Monoculture Robust Learning
We introduce a notion of monoculture robustness in Probably Approximately Correct (PAC) learning, as a way to mitigate the problem of algorithmic monoculture (Kleinberg and Raghavan, 2021): the tendency of models sharing the same components (e.g., training data) to fail on the same inputs. Specifically, we ask whether a learner with access to a single dataset can output individually accurate hypotheses whose joint failure probability is comparable to training on independent datasets. We say that such a learning procedure achieves monoculture robustness, and we quantify this by bounding the conjunctive error that all hypotheses misclassify points in an individual sense. We design methods for achieving this type of guarantee by drawing a surprising connection between replicability (Impagliazzo et al., 2022) and monoculture robustness. Specifically, we show that if the base learner is -replicable, then running it independently times on the same sample yields individual monoculture error that shrinks exponentially with , controlled by the single-run error of the base learner at the point and . We complement these positive results with a lower bound: for individual monoculture robustness, we prove a worst-case lower bound showing that, for learners, driving the individual conjunctive error down to can require a sample-size multiplier of .
查看此版本解读 →查看此版本原文 ↗