最近有很多不可能的结果表明，在与对抗对手的马尔可夫游戏中最小化的遗憾在统计学上和计算上是棘手的。然而，这些结果都没有排除在所有各方采用相同学习程序的假设下，遗憾最小化的可能性。在这项工作中，我们介绍了第一种（据我们所知）在通用马尔可夫游戏中学习的算法，该算法在所有代理商执行时提供了sublinear后悔保证。我们获得的边界是为了置换遗憾，因此，在此过程中，意味着融合了相关的平衡。我们的算法是分散的，计算上有效的，并且不需要代理之间的任何通信。我们的主要观察结果是，在马尔可夫游戏中通过策略优化的在线学习基本上减少了一种加权遗憾的最小化形式，而未知权重由代理商的策略顺序的路径长度确定。因此，控制路径长度会导致加权的遗憾目标，以提供足够的适应性算法提供统一的后悔保证。

我们研究随机梯度下降（SGD）在多大程度上被理解为“常规”学习规则，该规则通过获得良好的培训数据来实现概括性能。我们考虑基本的随机凸优化框架，其中（一通道，无需替代）SGD在经典上是众所周知的，可以最大程度地降低人口风险，以$ o（1/\ sqrt n）$ $ O（1/\ sqrt n）$，并且出人意料地证明，存在问题实例SGD解决方案既表现出$ \ omega（1）$的经验风险和概括差距。因此，事实证明，从任何意义上讲，SGD在算法上都不是稳定的，并且其概括能力不能通过均匀的收敛性或任何其他当前已知的概括性结合技术来解释（除了其经典分析外）。然后，我们继续分析与替代SGD密切相关的相关性，为此我们表明不会发生类似现象，并证明其人口风险实际上确实以最佳速度融合。最后，我们在没有替换SGD的背景下解释了我们的主要结果，用于有限的和凸优化问题，并得出多上类别制度的上限和下限，从而在先前已知的结果上有了显着改善。

Although most reinforcement learning research has centered on competitive games, little work has been done on applying it to co-operative multiplayer games or text-based games. Codenames is a board game that involves both asymmetric co-operation and natural language processing, which makes it an excellent candidate for advancing RL research. To my knowledge, this work is the first to formulate Codenames as a Markov Decision Process and apply some well-known reinforcement learning algorithms such as SAC, PPO, and A2C to the environment. Although none of the above algorithms converge for the Codenames environment, neither do they converge for a simplified environment called ClickPixel, except when the board size is small.

Multilingual machine translation models can benefit from synergy between different language pairs, but also suffer from interference. While there is a growing number of sophisticated methods that aim to eliminate interference, our understanding of interference as a phenomenon is still limited. This work identifies the main factors that contribute to interference in multilingual machine translation. Through systematic experimentation, we find that interference (or synergy) are primarily determined by model size, data size, and the proportion of each language pair within the total dataset. We observe that substantial interference occurs mainly when the model is very small with respect to the available training data, and that using standard transformer configurations with less than one billion parameters largely alleviates interference and promotes synergy. Moreover, we show that tuning the sampling temperature to control the proportion of each language pair in the data is key to balancing the amount of interference between low and high resource language pairs effectively, and can lead to superior performance overall.

We construct a universally Bayes consistent learning rule that satisfies differential privacy (DP). We first handle the setting of binary classification and then extend our rule to the more general setting of density estimation (with respect to the total variation metric). The existence of a universally consistent DP learner reveals a stark difference with the distribution-free PAC model. Indeed, in the latter DP learning is extremely limited: even one-dimensional linear classifiers are not privately learnable in this stringent model. Our result thus demonstrates that by allowing the learning rate to depend on the target distribution, one can circumvent the above-mentioned impossibility result and in fact, learn \emph{arbitrary} distributions by a single DP algorithm. As an application, we prove that any VC class can be privately learned in a semi-supervised setting with a near-optimal \emph{labeled} sample complexity of $\tilde{O}(d/\varepsilon)$ labeled examples (and with an unlabeled sample complexity that can depend on the target distribution).

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.

The problem of learning threshold functions is a fundamental one in machine learning. Classical learning theory implies sample complexity of $O(\xi^{-1} \log(1/\beta))$ (for generalization error $\xi$ with confidence $1-\beta$). The private version of the problem, however, is more challenging and in particular, the sample complexity must depend on the size $|X|$ of the domain. Progress on quantifying this dependence, via lower and upper bounds, was made in a line of works over the past decade. In this paper, we finally close the gap for approximate-DP and provide a nearly tight upper bound of $\tilde{O}(\log^* |X|)$, which matches a lower bound by Alon et al (that applies even with improper learning) and improves over a prior upper bound of $\tilde{O}((\log^* |X|)^{1.5})$ by Kaplan et al. We also provide matching upper and lower bounds of $\tilde{\Theta}(2^{\log^*|X|})$ for the additive error of private quasi-concave optimization (a related and more general problem). Our improvement is achieved via the novel Reorder-Slice-Compute paradigm for private data analysis which we believe will have further applications.

Segmentation of regions of interest (ROIs) for identifying abnormalities is a leading problem in medical imaging. Using Machine Learning (ML) for this problem generally requires manually annotated ground-truth segmentations, demanding extensive time and resources from radiologists. This work presents a novel weakly supervised approach that utilizes binary image-level labels, which are much simpler to acquire, to effectively segment anomalies in medical Magnetic Resonance (MR) images without ground truth annotations. We train a binary classifier using these labels and use it to derive seeds indicating regions likely and unlikely to contain tumors. These seeds are used to train a generative adversarial network (GAN) that converts cancerous images to healthy variants, which are then used in conjunction with the seeds to train a ML model that generates effective segmentations. This method produces segmentations that achieve Dice coefficients of 0.7903, 0.7868, and 0.7712 on the MICCAI Brain Tumor Segmentation (BraTS) 2020 dataset for the training, validation, and test cohorts respectively. We also propose a weakly supervised means of filtering the segmentations, removing a small subset of poorer segmentations to acquire a large subset of high quality segmentations. The proposed filtering further improves the Dice coefficients to up to 0.8374, 0.8232, and 0.8136 for training, validation, and test, respectively.

Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and applications using LLMs, we publicly release our models and code under the Responsible AI License.

We address the general task of structured commonsense reasoning: given a natural language input, the goal is to generate a graph such as an event -- or a reasoning-graph. To employ large language models (LMs) for this task, existing approaches ``serialize'' the output graph as a flat list of nodes and edges. Although feasible, these serialized graphs strongly deviate from the natural language corpora that LMs were pre-trained on, hindering LMs from generating them correctly. In this paper, we show that when we instead frame structured commonsense reasoning tasks as code generation tasks, pre-trained LMs of code are better structured commonsense reasoners than LMs of natural language, even when the downstream task does not involve source code at all. We demonstrate our approach across three diverse structured commonsense reasoning tasks. In all these natural language tasks, we show that using our approach, a code generation LM (CODEX) outperforms natural-LMs that are fine-tuned on the target task (e.g., T5) and other strong LMs such as GPT-3 in the few-shot setting.