We study the ability of foundation models to learn representations for classification that are transferable to new, unseen classes. Recent results in the literature show that representations learned by a single classifier over many classes are competitive on few-shot learning problems with representations learned by special-purpose algorithms designed for such problems. We offer an explanation for this phenomenon based on the concept of class-features variability collapse, which refers to the training dynamics of deep classification networks where the feature embeddings of samples belonging to the same class tend to concentrate around their class means. More specifically, we examine the few-shot error of the learned feature map, which is the classification error of the nearest class-center classifier using centers learned from a small number of random samples from each class. Assuming that the classes appearing in the data are selected independently from a distribution, we show that the few-shot error generalizes from the training data to unseen test data, and we provide an upper bound on the expected few-shot error for new classes (selected from the same distribution) using the average few-shot error for the source classes. Additionally, we show that the few-shot error on the training data can be upper bounded using the degree of class-features variability collapse. This suggests that foundation models can provide feature maps that are transferable to new downstream tasks even with limited data available.

We study the learning dynamics of self-predictive learning for reinforcement learning, a family of algorithms that learn representations by minimizing the prediction error of their own future latent representations. Despite its recent empirical success, such algorithms have an apparent defect: trivial representations (such as constants) minimize the prediction error, yet it is obviously undesirable to converge to such solutions. Our central insight is that careful designs of the optimization dynamics are critical to learning meaningful representations. We identify that a faster paced optimization of the predictor and semi-gradient updates on the representation, are crucial to preventing the representation collapse. Then in an idealized setup, we show self-predictive learning dynamics carries out spectral decomposition on the state transition matrix, effectively capturing information of the transition dynamics. Building on the theoretical insights, we propose bidirectional self-predictive learning, a novel self-predictive algorithm that learns two representations simultaneously. We examine the robustness of our theoretical insights with a number of small-scale experiments and showcase the promise of the novel representation learning algorithm with large-scale experiments.

We study distributed contextual linear bandits with stochastic contexts, where $N$ agents act cooperatively to solve a linear bandit-optimization problem with $d$-dimensional features over the course of $T$ rounds. For this problem, we derive the first ever information-theoretic lower bound $\Omega(dN)$ on the communication cost of any algorithm that performs optimally in a regret minimization setup. We then propose a distributed batch elimination version of the LinUCB algorithm, DisBE-LUCB, where the agents share information among each other through a central server. We prove that the communication cost of DisBE-LUCB matches our lower bound up to logarithmic factors. In particular, for scenarios with known context distribution, the communication cost of DisBE-LUCB is only $\tilde{\mathcal{O}}(dN)$ and its regret is ${\tilde{\mathcal{O}}}(\sqrt{dNT})$, which is of the same order as that incurred by an optimal single-agent algorithm for $NT$ rounds. We also provide similar bounds for practical settings where the context distribution can only be estimated. Therefore, our proposed algorithm is nearly minimax optimal in terms of \emph{both regret and communication cost}. Finally, we propose DecBE-LUCB, a fully decentralized version of DisBE-LUCB, which operates without a central server, where agents share information with their \emph{immediate neighbors} through a carefully designed consensus procedure.

我们研究了一个顺序决策问题，其中学习者面临$ k $武装的随机匪徒任务的顺序。对手可能会设计任务，但是对手受到限制，以在$ m $ and的较小（但未知）子集中选择每个任务的最佳组。任务边界可能是已知的（强盗元学习设置）或未知（非平稳的强盗设置）。我们设计了一种基于Burnit subsodular最大化的减少的算法，并表明，在大量任务和少数最佳武器的制度中，它在两种情况下的遗憾都比$ \ tilde {o}的简单基线要小。 \ sqrt {knt}）$可以通过使用为非平稳匪徒问题设计的标准算法获得。对于固定任务长度$ \ tau $的强盗元学习问题，我们证明该算法的遗憾被限制为$ \ tilde {o}（nm \ sqrt {m \ tau}+n^{2/3} m \ tau）$。在每个任务中最佳武器的可识别性的其他假设下，我们显示了一个带有改进的$ \ tilde {o}（n \ sqrt {m \ tau}+n^{1/2} {1/2} \ sqrt的强盗元学习算法{m k \ tau}）$遗憾。

我们研究了基础模型的能力，以了解可转让给新的看不见的课程的分类的表现。文献中最近的结果表明，单个分类器在许多课程中学到的表示在少量学习问题上具有竞争力，这些问题是由专为这些问题设计的特殊用途算法学习的表示。在本文中，我们基于最近观察到的现象提供了对这种行为的解释，即通过共同计量的分类网络学习的特征显示有趣的聚类属性，称为神经崩溃。理论上，我们在理论上展示了神经崩溃的展示给来自培训类的新样本，更重要的是 - 对于新课程，允许基础模型提供在转移学习中良好工作的特征地图，具体地，少量拍摄设置。

Dyadic and small group collaboration is an evolutionary advantageous behaviour and the need for such collaboration is a regular occurrence in day to day life. In this paper we estimate the perceived personality traits of individuals in dyadic and small groups over thin-slices of interaction on four multimodal datasets. We find that our transformer based predictive model performs similarly to human annotators tasked with predicting the perceived big-five personality traits of participants. Using this model we analyse the estimated perceived personality traits of individuals performing tasks in small groups and dyads. Permutation analysis shows that in the case of small groups undergoing collaborative tasks, the perceived personality of group members clusters, this is also observed for dyads in a collaborative problem solving task, but not in dyads under non-collaborative task settings. Additionally, we find that the group level average perceived personality traits provide a better predictor of group performance than the group level average self-reported personality traits.

Property inference attacks against machine learning (ML) models aim to infer properties of the training data that are unrelated to the primary task of the model, and have so far been formulated as binary decision problems, i.e., whether or not the training data have a certain property. However, in industrial and healthcare applications, the proportion of labels in the training data is quite often also considered sensitive information. In this paper we introduce a new type of property inference attack that unlike binary decision problems in literature, aim at inferring the class label distribution of the training data from parameters of ML classifier models. We propose a method based on \emph{shadow training} and a \emph{meta-classifier} trained on the parameters of the shadow classifiers augmented with the accuracy of the classifiers on auxiliary data. We evaluate the proposed approach for ML classifiers with fully connected neural network architectures. We find that the proposed \emph{meta-classifier} attack provides a maximum relative improvement of $52\%$ over state of the art.

To apply federated learning to drug discovery we developed a novel platform in the context of European Innovative Medicines Initiative (IMI) project MELLODDY (grant n{\deg}831472), which was comprised of 10 pharmaceutical companies, academic research labs, large industrial companies and startups. The MELLODDY platform was the first industry-scale platform to enable the creation of a global federated model for drug discovery without sharing the confidential data sets of the individual partners. The federated model was trained on the platform by aggregating the gradients of all contributing partners in a cryptographic, secure way following each training iteration. The platform was deployed on an Amazon Web Services (AWS) multi-account architecture running Kubernetes clusters in private subnets. Organisationally, the roles of the different partners were codified as different rights and permissions on the platform and administrated in a decentralized way. The MELLODDY platform generated new scientific discoveries which are described in a companion paper.

监督的机器学习为各种计算机视觉问题提供了最新的解决方案。但是，对大量标记的培训数据的需求限制了这些算法在稀缺或昂贵的情况下的这些算法的功能。自我监督的学习提供了一种方法，可以通过对未标记数据的特定域进行预处理模型来降低对手动注释数据的需求。在这种方法中，标记的数据完全需要用于微调下游任务的模型。医疗图像细分是一个标签数据需要专家知识并收集大型标记数据集的领域。因此，自我监督的学习算法有望在该领域进行实质性改进。尽管如此，自我监督的学习算法很少用于预识医学图像分割网络。在本文中，我们详细阐述并分析了对下游医学图像分割的监督和自我监督预审方法的有效性，重点是收敛和数据效率。我们发现，对自然图像和目标域特异性图像进行自我监督的预测会导致最快，最稳定的下游收敛性。在我们对ACDC心脏分割数据集的实验中，与Imagenet预处理的模型相比，这种预处理的方法可实现4-5倍的微调收敛。我们还表明，这种方法需要在域特异性数据上进行少于五个时期的预处理，以在下游收敛时间进行这种改进。最后，我们发现，在低数据方案中，有监督的Imagenet预处理达到了最佳准确性，需要少于100个带注释的样品才能实现接近最小误差。

作为人类已知的最直观的界面之一，自然语言有可能调解许多涉及人类计算机互动的任务，尤其是在音乐信息检索等以应用程序为中心的领域。在这项工作中，我们探索了跨模式学习，以试图在音乐领域弥合音频和语言。为此，我们提出了Muscall，这是音乐对比的音频学习框架。我们的方法由双重编码架构组成，该体系结构了解音乐音频和描述性句子对之间的对齐方式，生成可用于文本到原告和音频到文本检索的多模式嵌入。多亏了这个属性，肌肉几乎可以转移到任何可以作为基于文本检索的任务转移到任何任务。我们的实验表明，我们的方法在检索音频时的性能要比基线要好得多，该音频与文本描述匹配，相反，与音频查询匹配的文本。我们还证明，我们的模型的多模式对齐能力可以成功扩展到零摄像转移方案，用于流派分类和在两个公共数据集上自动标记。