On characterizations of learnability with computable learners
Munich Center for Mathematical Philosophy, LMU Munich（慕尼黑大学慕尼黑数学哲学中心）；CWI Amsterdam（阿姆斯特丹国家数学与计算机科学研究所）
Tom F. Sterkenburg
备注：
Final version, as accepted for COLT 2022. Changes w.r.t. previous arXiv version: In response to reviewer comments, major revision of the discussion in Section 3.2, incl. reformulation of Question 2 (that in its original form had an easy answer) and retraction of Definition 6 and Proposition 2 (that lost their relevance in the revised discussion). Minor textual revisions elsewhere
发表信息：
Proceedings of Thirty Fifth Conference on Learning Theory (COLT 2022), PMLR 178:3365-3379, 2022
We study computable PAC (CPAC) learning as introduced by Agarwal et al. (2020). First, we consider the main open question of finding characterizations of proper and improper CPAC learning. We give a characterization of a closely related notion of strong CPAC learning, and provide a negative answer to the COLT open problem posed by Agarwal et al. (2021) whether all decidably representable VC classes are improperly CPAC learnable. Second, we consider undecidability of (computable) PAC learnability. We give a simple general argument to exhibit such ndecidability, and initiate a study of the arithmetical complexity of learnability. We briefly discuss the relation to the undecidability result of Ben-David et al. (2019), that motivated the work of Agarwal et al.
On the Quantum versus Classical Learnability of Discrete Distributions
Dahlem Center for Complex Quantum Systems, Freie Universität Berlin（柏林自由大学达勒姆复杂量子系统中心）；TU Berlin（柏林工业大学）；FhG SIT（弗劳恩霍夫系统与创新研究所）；Helmholtz Center Berlin（柏林亥姆霍兹中心）；Department of Mathematics and Computer Science, Freie Universität Berlin（柏林自由大学数学与计算机科学系）
Physics-Aware Learnability: From Set-Theoretic Independence to Operational Constraints
Institute for Convergence Research and Education in Advanced Technology（融合研究中心暨先进技术教育研究所）；Yonsei University（延世大学）；Department of Quantum Information（量子信息系）；Department of Statistics and Data Science（统计与数据科学系）
Deep pNML: Predictive Normalized Maximum Likelihood for Deep Neural Networks
Tel Aviv University（特拉维夫大学）
预测归一化最大似然估计（Predictive Normalized Maximum Likewise，pNML）方案最近被建议用于个体设置中的通用学习，其中训练和测试样本都是个体数据。通用学习的目标是与知道数据值的“genie”或参考学习器竞争，但限于使用来自给定模型类的学习器。pNML最小化未知标签的任何可能值的相关遗憾。此外，它的最小-最大遗憾可以作为特定训练和数据样本的可学习性的逐点度量。在这项工作中，我们研究了深度神经网络（DNN）模型类的pNML及其相关的可学习性度量。如图所示，pNML优于常用的经验风险最小化（ERM）方法，并提供了对对抗攻击的鲁棒性。连同其可学习性度量，它可以检测出分布测试示例，容忍噪声标签，并作为ERM的置信度度量。最后，我们将pNML扩展到一个“两次通用”的解决方案，该解决方案为模型类选择提供了通用性，并生成了一个与所有模型类中最好的学习者竞争的学习者。
Recursively Enumerably Representable Classes and Computable Versions of the Fundamental Theorem of Statistical Learning
Institute for Symbolic Artificial Intelligence, Johannes Kepler University Linz（约翰内斯·开普勒大学林茨分校符号人工智能研究所）；Institute for the Interdisciplinary Study of Language Evolution, University of Zurich（苏黎世大学语言演化跨学科研究所）；Department of Mathematics and Statistics, University of Konstanz（康斯坦茨大学数学与统计学系）
Find a witness or shatter: the landscape of computable PAC learning
Centro Nacional de Inteligencia Artificial（国家人工智能中心）；Instituto de Ingeniería Matemática y Computacional, Universidad Católica de Chile（智利天主教大学数学与计算工程学院）；Instituto Milenio Fundamentos de los Datos（数据基础千年研究所）；Institute of Fundamental Technological Research, Polish Academy of Sciences（波兰科学院基础技术研究所）
Sapienza University of Rome（罗马大学）；Centro Nacional de Inteligencia Artificial（国家人工智能中心）；Universidad Católica de Chile（智利天主教大学）；Institute of Fundamental Technological Research, Polish Academy of Sciences（波兰科学院基础技术研究所）
