Generalized Chinese Remainder Theorem (CRT) has been shown to be a powerful approach to solve the ambiguity resolution problem. However, with its close relationship to number theory, study in this area is mainly from a coding theory perspective under deterministic conditions. Nevertheless, it can be proved that even with the best deterministic condition known, the probability of success in robust reconstruction degrades exponentially as the number of estimand increases. In this paper, we present the first rigorous analysis on the underlying statistical model of CRT-based multiple parameter estimation, where a generalized Gaussian mixture with background knowledge on samplings is proposed. To address the problem, two novel approaches are introduced. One is to directly calculate the conditional maximal a posteriori probability (MAP) estimation of residue clustering, and the other is to iteratively search for MAP of both common residues and clustering. Moreover, remainder error-correcting codes are introduced to improve the robustness further. It is shown that this statistically based scheme achieves much stronger robustness compared to state-of-the-art deterministic schemes, especially in low and median Signal Noise Ratio (SNR) scenarios.

This paper studies distributed estimation and inference for a general statistical problem with a convex loss that could be non-differentiable. For the purpose of efficient computation, we restrict ourselves to stochastic first-order optimization, which enjoys low per-iteration complexity. To motivate the proposed method, we first investigate the theoretical properties of a straightforward Divide-and-Conquer Stochastic Gradient Descent (DC-SGD) approach. Our theory shows that there is a restriction on the number of machines and this restriction becomes more stringent when the dimension $p$ is large. To overcome this limitation, this paper proposes a new multi-round distributed estimation procedure that approximates the Newton step only using stochastic subgradient. The key component in our method is the proposal of a computationally efficient estimator of $\Sigma^{-1} w$, where $\Sigma$ is the population Hessian matrix and $w$ is any given vector. Instead of estimating $\Sigma$ (or $\Sigma^{-1}$) that usually requires the second-order differentiability of the loss, the proposed First-Order Newton-type Estimator (FONE) directly estimates the vector of interest $\Sigma^{-1} w$ as a whole and is applicable to non-differentiable losses. Our estimator also facilitates the inference for the empirical risk minimizer. It turns out that the key term in the limiting covariance has the form of $\Sigma^{-1} w$, which can be estimated by FONE.

An accurate model of a patient's individual survival distribution can help determine the appropriate treatment for terminal patients. Unfortunately, risk scores (e.g., from Cox Proportional Hazard models) do not provide survival probabilities, single-time probability models (e.g., the Gail model, predicting 5 year probability) only provide for a single time point, and standard Kaplan-Meier survival curves provide only population averages for a large class of patients meaning they are not specific to individual patients. This motivates an alternative class of tools that can learn a model which provides an individual survival distribution which gives survival probabilities across all times - such as extensions to the Cox model, Accelerated Failure Time, an extension to Random Survival Forests, and Multi-Task Logistic Regression. This paper first motivates such "individual survival distribution" (ISD) models, and explains how they differ from standard models. It then discusses ways to evaluate such models - namely Concordance, 1-Calibration, Brier score, and various versions of L1-loss - and then motivates and defines a novel approach "D-Calibration", which determines whether a model's probability estimates are meaningful. We also discuss how these measures differ, and use them to evaluate several ISD prediction tools, over a range of survival datasets.

Conditional GANs are at the forefront of natural image synthesis. The main drawback of such models is the necessity for labelled data. In this work we exploit two popular unsupervised learning techniques, adversarial training and self-supervision, to close the gap between conditional and unconditional GANs. In particular, we allow the networks to collaborate on the task of representation learning, while being adversarial with respect to the classic GAN game. The role of self-supervision is to encourage the discriminator to learn meaningful feature representations which are not forgotten during training. We test empirically both the quality of the learned image representations, and the quality of the synthesized images. Under the same conditions, the self-supervised GAN attains a similar performance to state-of-the-art conditional counterparts. Finally, we show that this approach to fully unsupervised learning can be scaled to attain an FID of 33 on unconditional ImageNet generation.

Model distillation aims to distill the knowledge of a complex model into a simpler one. In this paper, we consider an alternative formulation called {\em dataset distillation}: we keep the model fixed and instead attempt to distill the knowledge from a large training dataset into a small one. The idea is to {\em synthesize} a small number of data points that do not need to come from the correct data distribution, but will, when given to the learning algorithm as training data, approximate the model trained on the original data. For example, we show that it is possible to compress $60,000$ MNIST training images into just $10$ synthetic {\em distilled images} (one per class) and achieve close to original performance with only a few steps of gradient descent, given a particular fixed network initialization. We evaluate our method in a wide range of initialization settings and with different learning objectives. Experiments on multiple datasets show the advantage of our approach compared to alternative methods in most settings.

Conventional surveillance systems for monitoring infectious diseases, such as influenza, face challenges due to shortage of skilled healthcare professionals, remoteness of communities and absence of communication infrastructures. Internet-based approaches for surveillance are appealing logistically as well as economically. Search engine queries and Twitter have been the primarily used data sources in such approaches. The aim of this study is to assess the predictive power of an alternative data source, Instagram. By using 317 weeks of publicly available data from Instagram, we trained several machine learning algorithms to both nowcast and forecast the number of official influenza-like illness incidents in Finland where population-wide official statistics about the weekly incidents are available. In addition to date and hashtag count features of online posts, we were able to utilize also the visual content of the posted images with the help of deep convolutional neural networks. Our best nowcasting model reached a mean absolute error of 11.33 incidents per week and a correlation coefficient of 0.963 on the test data. Forecasting models for predicting 1 week and 2 weeks ahead showed statistical significance as well by reaching correlation coefficients of 0.903 and 0.862, respectively. This study demonstrates how social media and in particular, digital photographs shared in them, can be a valuable source of information for the field of infodemiology.

We analyze speed of convergence to global optimum for gradient descent training a deep linear neural network (parameterized as $x \mapsto W_N W_{N-1} \cdots W_1 x$) by minimizing the $\ell_2$ loss over whitened data. Convergence at a linear rate is guaranteed when the following hold: (i) dimensions of hidden layers are at least the minimum of the input and output dimensions; (ii) weight matrices at initialization are approximately balanced; and (iii) the initial loss is smaller than the loss of any rank-deficient solution. The assumptions on initialization (conditions (ii) and (iii)) are necessary, in the sense that violating any one of them may lead to convergence failure. Moreover, in the important case of output dimension 1, i.e. scalar regression, they are met, and thus convergence to global optimum holds, with constant probability under a random initialization scheme. Our results significantly extend previous analyses, e.g., of deep linear residual networks (Bartlett et al., 2018).

Dimensionality reduction is a common method for analyzing and visualizing high-dimensional data. However, reasoning dynamically about the results of a dimensionality reduction is difficult. Dimensionality-reduction algorithms use complex optimizations to reduce the number of dimensions of a dataset, but these new dimensions often lack a clear relation to the initial data dimensions, thus making them difficult to interpret. Here we propose a visual interaction framework to improve dimensionality-reduction based exploratory data analysis. We introduce two interaction techniques, forward projection and backward projection, for dynamically reasoning about dimensionally reduced data. We also contribute two visualization techniques, prolines and feasibility maps, to facilitate the effective use of the proposed interactions. We apply our framework to PCA and autoencoder-based dimensionality reductions. Through data-exploration examples, we demonstrate how our visual interactions can improve the use of dimensionality reduction in exploratory data analysis.

This technical report describes the multi-label classification (MLC) search space in the MEKA software, including the traditional/meta MLC algorithms, and the traditional/meta/pre-processing single-label classification (SLC) algorithms. The SLC search space is also studied because is part of MLC search space as several methods use problem transformation methods to create a solution (i.e., a classifier) for a MLC problem. This was done in order to understand better the MLC algorithms. Finally, we propose a grammar that formally expresses this understatement.

Variational inference (VI) has become the method of choice for fitting many modern probabilistic models. However, practitioners are faced with a fragmented literature that offers a bewildering array of algorithmic options. First, the variational family. Second, the granularity of the updates e.g. whether the updates are local to each data point and employ message passing or global. Third, the method of optimization (bespoke or blackbox, closed-form or stochastic updates, etc.). This paper presents a new framework, termed Partitioned Variational Inference (PVI), that explicitly acknowledges these algorithmic dimensions of VI, unifies disparate literature, and provides guidance on usage. Crucially, the proposed PVI framework allows us to identify new ways of performing VI that are ideally suited to challenging learning scenarios including federated learning (where distributed computing is leveraged to process non-centralized data) and continual learning (where new data and tasks arrive over time and must be accommodated quickly). We showcase these new capabilities by developing communication-efficient federated training of Bayesian neural networks and continual learning for Gaussian process models with private pseudo-points. The new methods significantly outperform the state-of-the-art, whilst being almost as straightforward to implement as standard VI.