The over-parameterized models attract much attention in the era of data science and deep learning. It is empirically observed that although these models, e.g. deep neural networks, over-fit the training data, they can still achieve small testing error, and sometimes even {\em outperform} traditional algorithms which are designed to avoid over-fitting. The major goal of this work is to sharply quantify the benefit of data interpolation in the context of nearest neighbors (NN) algorithm. Specifically, we consider a class of interpolated weighting schemes and then carefully characterize their asymptotic performances. Our analysis reveals a U-shaped performance curve with respect to the level of data interpolation, and proves that a mild degree of data interpolation {\em strictly} improves the prediction accuracy and statistical stability over those of the (un-interpolated) optimal $k$NN algorithm. This theoretically justifies (predicts) the existence of the second U-shaped curve in the recently discovered double descent phenomenon. Note that our goal in this study is not to promote the use of interpolated-NN method, but to obtain theoretical insights on data interpolation inspired by the aforementioned phenomenon.

Quantifying the value of data is a fundamental problem in machine learning. Data valuation has multiple important use cases: (1) building insights about the learning task, (2) domain adaptation, (3) corrupted sample discovery, and (4) robust learning. To adaptively learn data values jointly with the target task predictor model, we propose a meta learning framework which we name Data Valuation using Reinforcement Learning (DVRL). We employ a data value estimator (modeled by a deep neural network) to learn how likely each datum is used in training of the predictor model. We train the data value estimator using a reinforcement signal of the reward obtained on a small validation set that reflects performance on the target task. We demonstrate that DVRL yields superior data value estimates compared to alternative methods across different types of datasets and in a diverse set of application scenarios. The corrupted sample discovery performance of DVRL is close to optimal in many regimes (i.e. as if the noisy samples were known apriori), and for domain adaptation and robust learning DVRL significantly outperforms state-of-the-art by 14.6% and 10.8%, respectively.

In this paper, we consider data consisting of multiple networks, each comprised of a different edge set on a common set of nodes. Many models have been proposed for such multi-view data, assuming that the data views are closely related. In this paper, we provide tools for evaluating the assumption that there is a relationship between the different views. In particular, we ask: is there an association between the latent community memberships of the nodes within each data view? To answer this question, we extend the stochastic block model for a single network view to two network views, and develop a new hypothesis test for the null hypothesis that the latent community structure within each data view is independent. We apply our test to protein-protein interaction data sets from the HINT database (Das & Yu 2012). We find evidence of a weak association between the latent community structure of proteins defined with respect to binary interaction data and with respect to co-complex association data. We also extend this proposal to the setting of a network with node covariates.

A BSTRACT The Dynamical Gaussian Process Latent V ariable Models provide an elegant nonparametric framework for learning the low dimensional representations of the high-dimensional time-series. Real world observational studies, however, are often ill-conditioned: the observations can be noisy, not assuming the luxury of relatively complete and equally spaced like those in time series. Such conditions make it difficult to learn reasonable representations in the high dimensional longitudinal data set by way of Gaussian Process Latent V ariable Model as well as other dimensionality reduction procedures. In this study, we approach the inference of Gaussian Process Dynamical Systems in Longitudinal scenario by augmenting the bound in the variational approximation to include systematic samples of the unseen observations. We demonstrate the usefulness of this approach on synthetic as well as the human motion capture data set. 1 I NTRODUCTION While it isn't trivial to find an unified definition of multivariate longitudinal data; the one definition being alluded to in Pullenayegum & Lim (2016) is the type of data being discussed in this work. Longitudinal designs track a repeated set of variables in experimental subjects over periods of time; however, unlike time series, which are often characterized by regular intervals of n-dimensional observations, longitudinal setups present inconsistent sampling frequencies and only a small subset of variables may be observed at any given time.

Many methods have been developed to represent knowledge graph data, which implicitly exploit low-rank latent structure in the data to encode known information and enable unknown facts to be inferred. To predict whether a relationship holds between entities, their embeddings are typically compared in the latent space following a relation-specific mapping. Whilst link prediction has steadily improved, the latent structure, and hence why such models capture semantic information, remains unexplained. We build on recent theoretical interpretation of word embeddings as a basis to consider an explicit structure for representations of relations between entities. For identifiable relation types, we are able to predict properties and justify the relative performance of leading knowledge graph representation methods, including their often overlooked ability to make independent predictions.

Deep Graph Neural Networks (GNNs) are instrumental in graph classification and graph-based regression tasks. In these tasks, graph pooling is a critical ingredient by which GNNs adapt to input graphs of varying size and structure. We propose a new graph pooling operation based on compressive Haar transforms, called HaarPooling. HaarPooling is computed following a chain of sequential clusterings of the input graph. The input of each pooling layer is transformed by the compressive Haar basis of the corresponding clustering. HaarPooling operates in the frequency domain by the synthesis of nodes in the same cluster and filters out fine detail information by compressive Haar transforms. Such transforms provide an effective characterization of the data and preserve the structure information of the input graph. By the sparsity of the Haar basis, the computation of HaarPooling is of linear complexity. The GNN with HaarPooling and existing graph convolution layers achieves state-of-the-art performance on diverse graph classification problems.

It has been widely recognized that adversarial examples can be easily crafted to fool deep networks, which mainly root from the locally non-linear behavior nearby input examples. Applying mixup in training provides an effective mechanism to improve generalization performance and model robustness against adversarial perturbations, which introduces the globally linear behavior in-between training examples. However, in previous work, the mixup-trained models only passively defend adversarial attacks in inference by directly classifying the inputs, where the induced global linearity is not well exploited. Namely, since the locality of the adversarial perturbations, it would be more efficient to actively break the locality via the globality of the model predictions. Inspired by simple geometric intuition, we develop an inference principle, named mixup inference (MI), for mixup-trained models. MI mixups the input with other random clean samples, which can shrink and transfer the equivalent perturbation if the input is adversarial. Our experiments on CIFAR-10 and CIFAR-100 demonstrate that MI can further improve the adversarial robustness for the models trained by mixup and its variants.

In clustering we normally output one cluster variable for each datapoint. However it is not necessarily the case that there is only one way to partition a given dataset into cluster components. For example, one could cluster objects by their colour, or by their type. Different attributes form a hierarchy, and we could wish to cluster in any of them. By disentangling the learnt latent representations of some dataset into different layers for different attributes we can then cluster in those latent spaces. We call this "disentangled clustering". Extending Variational Ladder Autoencoders (Zhao et al., 2017), we propose a clustering algorithm, VLAC, that outperforms a Gaussian Mixture DGM in cluster accuracy over digit identity on the test set of SVHN. We also demonstrate learning clusters jointly over numerous layers of the hierarchy of latent variables for the data, and show component-wise generation from this hierarchical model.

Meta-learning has been widely used for implementing few-shot learning and fast model adaptation. One kind of meta-learning methods attempt to learn how to control the gradient descent process in order to make the gradient-based learning have high speed and generalization. This work proposes a method that controls the gradient descent process of the model parameters of a neural network by limiting the model parameters in a low-dimensional latent space. The main challenge of this idea is that a decoder with too many parameters is required. This work designs a decoder with typical structure and shares a part of weights in the decoder to reduce the number of the required parameters. Besides, this work has introduced ensemble learning to work with the proposed approach for improving performance. The results show that the proposed approach is witnessed by the superior performance over the Omniglot classification and the miniImageNet classification tasks.

In this paper, we study problems at the interface of two important fields: \emph{submodular optimization} and \emph{online learning}. Submodular functions play a vital role in modelling cost functions that naturally arise in many areas of discrete optimization. These functions have been studied under various models of computation. Independently, submodularity has been considered in continuous domains. In fact, many problems arising in machine learning and statistics have been modelled using continuous DR-submodular functions. In this work, we are study the problem of maximizing \textit{non-monotone} continuous DR-submodular functions within the framework of online learning. We provide three main results. First, we present an online algorithm (in full-information setting) that achieves an approximation guarantee (depending on the search space) for the problem of maximizing non-monotone continuous DR-submodular functions over a \emph{general} convex domain. To best of our knowledge, no prior approximation algorithm in full-information setting was known for the non-monotone continuous DR submodular functions even for the \emph{down-closed} convex domain. Second, we show that the online stochastic mirror ascent algorithm (in full information setting) achieves an improved approximation ratio of $(1/4)$ for maximizing the non-monotone continuous DR-submodular functions over a \emph{down-closed} convex domain. At last, we extend our second result to the bandit setting where we present the first approximation guarantee of $(1/4)$. To best of our knowledge, no approximation algorithm for non-monotone submodular maximization was known in the bandit setting.