在机器学习中，我们传统上评估单个模型的性能，平均在测试输入集合中进行平均。在这项工作中，我们提出了一种新方法：在$ \ textit {单个输入点} $上评估时，我们测量了模型集合的性能。具体来说，我们研究了一个点的$ \ textit {profile {profile} $：模型在测试分布上的平均性能与他们在该点上的角度表现之间的关系。我们发现配置文件可以在分布和分发的模型和数据的结构中产生新的见解。例如，我们从经验上表明，实际数据分布由具有质量不同的点组成。一方面，有“兼容”点，在角度和平均性能之间具有很强的相关性。另一方面，有些点具有弱甚至$ \ textit {nogate} $相关性：提高整体模型精度实际上$ \ textit {hurts} $性能的情况。我们证明，这些实验观察与先前工作中提出的几种简化学习模型的预测不一致。作为一个应用程序，我们使用配置文件来构造一个数据集，我们称为CIFAR-10-NENG：CINIC-10的子集，因此对于标准模型，CIFAR-10-NENG上的准确性为$ \ textit {negalissiper {negalissiperational {negalishatied} CIFAR-10测试。这首先说明了一个完全逆转“准确性”的OOD数据集（Miller，Taori，Raghunathan，Sagawa，Koh，Koh，Shankar，Liang，Carmon和Schmidt 2021）

We show that a variety of modern deep learning tasks exhibit a "double-descent" phenomenon where, as we increase model size, performance first gets worse and then gets better. Moreover, we show that double descent occurs not just as a function of model size, but also as a function of the number of training epochs. We unify the above phenomena by defining a new complexity measure we call the effective model complexity and conjecture a generalized double descent with respect to this measure. Furthermore, our notion of model complexity allows us to identify certain regimes where increasing (even quadrupling) the number of train samples actually hurts test performance. * Work performed in part while Preetum Nakkiran was interning at OpenAI, with Ilya Sutskever. We especially thank Mikhail Belkin and Christopher Olah for helpful discussions throughout this work.

KL-regularized reinforcement learning from expert demonstrations has proved successful in improving the sample efficiency of deep reinforcement learning algorithms, allowing them to be applied to challenging physical real-world tasks. However, we show that KL-regularized reinforcement learning with behavioral reference policies derived from expert demonstrations can suffer from pathological training dynamics that can lead to slow, unstable, and suboptimal online learning. We show empirically that the pathology occurs for commonly chosen behavioral policy classes and demonstrate its impact on sample efficiency and online policy performance. Finally, we show that the pathology can be remedied by non-parametric behavioral reference policies and that this allows KL-regularized reinforcement learning to significantly outperform state-of-the-art approaches on a variety of challenging locomotion and dexterous hand manipulation tasks.

State-of-the-art language models are often accurate on many question-answering benchmarks with well-defined questions. Yet, in real settings questions are often unanswerable without asking the user for clarifying information. We show that current SotA models often do not ask the user for clarification when presented with imprecise questions and instead provide incorrect answers or "hallucinate". To address this, we introduce CLAM, a framework that first uses the model to detect ambiguous questions, and if an ambiguous question is detected, prompts the model to ask the user for clarification. Furthermore, we show how to construct a scalable and cost-effective automatic evaluation protocol using an oracle language model with privileged information to provide clarifying information. We show that our method achieves a 20.15 percentage point accuracy improvement over SotA on a novel ambiguous question-answering answering data set derived from TriviaQA.

Learned classifiers should often possess certain invariance properties meant to encourage fairness, robustness, or out-of-distribution generalization. However, multiple recent works empirically demonstrate that common invariance-inducing regularizers are ineffective in the over-parameterized regime, in which classifiers perfectly fit (i.e. interpolate) the training data. This suggests that the phenomenon of ``benign overfitting," in which models generalize well despite interpolating, might not favorably extend to settings in which robustness or fairness are desirable. In this work we provide a theoretical justification for these observations. We prove that -- even in the simplest of settings -- any interpolating learning rule (with arbitrarily small margin) will not satisfy these invariance properties. We then propose and analyze an algorithm that -- in the same setting -- successfully learns a non-interpolating classifier that is provably invariant. We validate our theoretical observations on simulated data and the Waterbirds dataset.

Selecting subsets of features that differentiate between two conditions is a key task in a broad range of scientific domains. In many applications, the features of interest form clusters with similar effects on the data at hand. To recover such clusters we develop DiSC, a data-driven approach for detecting groups of features that differentiate between conditions. For each condition, we construct a graph whose nodes correspond to the features and whose weights are functions of the similarity between them for that condition. We then apply a spectral approach to compute subsets of nodes whose connectivity differs significantly between the condition-specific feature graphs. On the theoretical front, we analyze our approach with a toy example based on the stochastic block model. We evaluate DiSC on a variety of datasets, including MNIST, hyperspectral imaging, simulated scRNA-seq and task fMRI, and demonstrate that DiSC uncovers features that better differentiate between conditions compared to competing methods.

Graphons are general and powerful models for generating graphs of varying size. In this paper, we propose to directly model graphons using neural networks, obtaining Implicit Graphon Neural Representation (IGNR). Existing work in modeling and reconstructing graphons often approximates a target graphon by a fixed resolution piece-wise constant representation. Our IGNR has the benefit that it can represent graphons up to arbitrary resolutions, and enables natural and efficient generation of arbitrary sized graphs with desired structure once the model is learned. Furthermore, we allow the input graph data to be unaligned and have different sizes by leveraging the Gromov-Wasserstein distance. We first demonstrate the effectiveness of our model by showing its superior performance on a graphon learning task. We then propose an extension of IGNR that can be incorporated into an auto-encoder framework, and demonstrate its good performance under a more general setting of graphon learning. We also show that our model is suitable for graph representation learning and graph generation.

Offline reinforcement-learning (RL) algorithms learn to make decisions using a given, fixed training dataset without the possibility of additional online data collection. This problem setting is captivating because it holds the promise of utilizing previously collected datasets without any costly or risky interaction with the environment. However, this promise also bears the drawback of this setting. The restricted dataset induces subjective uncertainty because the agent can encounter unfamiliar sequences of states and actions that the training data did not cover. Moreover, inherent system stochasticity further increases uncertainty and aggravates the offline RL problem, preventing the agent from learning an optimal policy. To mitigate the destructive uncertainty effects, we need to balance the aspiration to take reward-maximizing actions with the incurred risk due to incorrect ones. In financial economics, modern portfolio theory (MPT) is a method that risk-averse investors can use to construct diversified portfolios that maximize their returns without unacceptable levels of risk. We integrate MPT into the agent's decision-making process to present a simple-yet-highly-effective risk-aware planning algorithm for offline RL. Our algorithm allows us to systematically account for the \emph{estimated quality} of specific actions and their \emph{estimated risk} due to the uncertainty. We show that our approach can be coupled with the Transformer architecture to yield a state-of-the-art planner for offline RL tasks, maximizing the return while significantly reducing the variance.

政策梯度方法被广泛用于学习控制政策。它们可以轻松地分配给多名工人，并在许多领域中达到最新结果。不幸的是，它们表现出很大的差异，随后遭受了高样本的复杂性，因为它们在整个轨迹上汇总了梯度。在另一个极端情况下，计划方法，例如树木搜索，使用考虑未来LookAhead的单步过渡来优化策略。这些方法主要用于基于价值的算法。基于计划的算法需要一个正向模型，并且在每个步骤上都是计算密集型的，但更有效。在这项工作中，我们介绍了SoftTreemax，这是将树搜索整合到策略梯度中的第一种方法。传统上，针对单个状态行动对计算梯度。取而代之的是，我们基于树的策略结构在每个环境步骤中利用树叶的所有梯度。这使我们能够将梯度的差异减少三个数量级，并与标准策略梯度相比，从更好的样本复杂性中受益。在Atari上，与分布式PPO相比，SoftTreemax在运行时的表现高达5倍。

培训低级的深层神经网络，即使用分解层，特别是社区感兴趣的：它在记忆消耗和训练时间方面提供了对未分离培训的效率。先前的工作集中在预训练的网络的低级近似值和低级空间中的培训中，并提供了其他目标，为所选实践提供了各种临时解释。我们分析了在实践中运作良好的技术，并通过对诸如GPT2之类的模型进行广泛的消融，我们提供了证据表明该领域的共同信念，这暗示着令人兴奋的研究机会仍然需要回答。