Various strategies for active learning have been proposed in the machine learning literature. In uncertainty sampling, which is among the most popular approaches, the active learner sequentially queries the label of those instances for which its current prediction is maximally uncertain. The predictions as well as the measures used to quantify the degree of uncertainty, such as entropy, are almost exclusively of a probabilistic nature. In this paper, we advocate a distinction between two different types of uncertainty, referred to as epistemic and aleatoric, in the context of active learning. Roughly speaking, these notions capture the reducible and the irreducible part of the total uncertainty in a prediction, respectively. We conjecture that, in uncertainty sampling, the usefulness of an instance is better reflected by its epistemic than by its aleatoric uncertainty. This leads us to suggest the principle of "epistemic uncertainty sampling", which we instantiate by means of a concrete approach for measuring epistemic and aleatoric uncertainty. In experimental studies, epistemic uncertainty sampling does indeed show promising performance.

The large volume of text in electronic healthcare records often remains underused due to a lack of methodologies to extract interpretable content. Here we present an unsupervised framework for the analysis of free text that combines text-embedding with paragraph vectors and graph-theoretical multiscale community detection. We analyse text from a corpus of patient incident reports from the National Health Service in England to find content-based clusters of reports in an unsupervised manner and at different levels of resolution. Our unsupervised method extracts groups with high intrinsic textual consistency and compares well against categories hand-coded by healthcare personnel. We also show how to use our content-driven clusters to improve the supervised prediction of the degree of harm of the incident based on the text of the report. Finally, we discuss future directions to monitor reports over time, and to detect emerging trends outside pre-existing categories.

State-of-the-art methods for Convolutional Sparse Coding usually employ Fourier-domain solvers in order to speed up the convolution operators. However, this approach is not without shortcomings. For example, Fourier-domain representations implicitly assume circular boundary conditions and make it hard to fully exploit the sparsity of the problem as well as the small spatial support of the filters. In this work, we propose a novel stochastic spatial-domain solver, in which a randomized subsampling strategy is introduced during the learning sparse codes. Afterwards, we extend the proposed strategy in conjunction with online learning, scaling the CSC model up to very large sample sizes. In both cases, we show experimentally that the proposed subsampling strategy, with a reasonable selection of the subsampling rate, outperforms the state-of-the-art frequency-domain solvers in terms of execution time without losing the learning quality. Finally, we evaluate the effectiveness of the over-complete dictionary learned from large-scale datasets, which demonstrates an improved sparse representation of the natural images on account of more abundant learned image features.

Machine Learning (ML), and especially Deep Learning (DL) methods, have evolved rapidly over the last years and showed remarkable advances in research areas such as computer vision and natural language processing; however, there are still engineering applications in industries such as transportation where DL methods have not been applied yet or that can be benefited from an integrated approach using DL in addition to other methods. For countries in Northern latitudes, one of such applications is Monitoring Road Surface Condition (RSC) during the Winter season for improving road safety and road maintenance operations. In this study, we introduce a novel approach for monitoring of RSC that integrates DL methods and Spatial Statistics (SS) to simultaneously process data from roadside cameras and weather stations to determine automatically the category of snow coverage at sample locations across a region of interest.

No matter how you prefer to learn, you have probably heard someone mention something about your particular "learning style." This seemingly simple phrase implies that there are many different ways to learn, and people can be better or worse at certain learning methods depending on personal preferences and circumstances. As it is with people, so we tend to make it for computers. Machine learning is moving forward at a fast rate thanks to researchers and programmers figuring out how to optimize the ways through which machines learn new information. Two methods in particular, called tree boosting and adversarial networks, work in tandem to produce encouraging results.

Algorithmic risk assessments are increasingly used to help humans make decisions in high-stakes settings, such as medicine, criminal justice and education. In each of these cases, the purpose of the risk assessment tool is to inform actions, such as medical treatments or release conditions, often with the aim of reducing the likelihood of an adverse event such as hospital readmission or recidivism. Problematically, most tools are trained and evaluated on historical data in which the outcomes observed depend on the historical decision-making policy. These tools thus reflect risk under the historical policy, rather than under the different decision options that the tool is intended to inform. Even when tools are constructed to predict risk under a specific decision, they are often improperly evaluated as predictors of the target outcome. Focusing on the evaluation task, in this paper we define counterfactual analogues of common predictive performance and algorithmic fairness metrics that we argue are better suited for the decision-making context. We introduce a new method for estimating the proposed metrics using doubly robust estimation. We provide theoretical results that show that only under strong conditions can fairness according to the standard metric and the counterfactual metric simultaneously hold. Consequently, fairness-promoting methods that target parity in a standard fairness metric may --- and as we show empirically, do --- induce greater imbalance in the counterfactual analogue. We provide empirical comparisons on both synthetic data and a real world child welfare dataset to demonstrate how the proposed method improves upon standard practice.

When the data is distributed across multiple servers, efficient data exchange between the servers (or workers) for solving the distributed learning problem is an important problem and is the focus of this paper. We propose a fast, privacy-aware, and communication-efficient decentralized framework to solve the distributed machine learning (DML) problem. The proposed algorithm, GADMM, is based on Alternating Direct Method of Multiplier (ADMM) algorithm. The key novelty in GADMM is that each worker exchanges the locally trained model only with two neighboring workers, thereby training a global model with lower amount of communication in each exchange. We prove that GADMM converges faster than the centralized batch gradient descent for convex loss functions, and numerically show that it is faster and more communication-efficient than the state-of-the-art communication-efficient centralized algorithms such as the Lazily Aggregated Gradient (LAG), in linear and logistic regression tasks on synthetic and real datasets. Furthermore, we propose Dynamic GADMM (D-GADMM), a variant of GADMM, and prove its convergence under time-varying network topology of the workers.

A versatile and effective approach to meta-learning is to infer a gradient-based up-date rule directly from data that promotes rapid learning of new tasks from the same distribution. Current methods rely on backpropagating through the learning process, limiting their scope to few-shot learning. In this work, we introduce Warped Gradient Descent (WarpGrad), a family of modular optimisers that can scale to arbitrary adaptation processes. WarpGrad methods meta-learn to warp task loss surfaces across the joint task-parameter distribution to facilitate gradient descent, which is achieved by a reparametrisation of neural networks that interleaves warp layers in the architecture. These layers are shared across task learners and fixed during adaptation; they represent a projection of task parameters into a meta-learned space that is conducive to task adaptation and standard backpropagation induces a form of gradient preconditioning. WarpGrad methods are computationally efficient and easy to implement as they rely on parameter sharing and backpropagation. They are readily combined with other meta-learners and can scale both in terms of model size and length of adaptation trajectories as meta-learning warp parameters do not require differentiation through task adaptation processes. We show empirically that WarpGrad optimisers meta-learn a warped space where gradient descent is well behaved, with faster convergence and better performance in a variety of settings, including few-shot, standard supervised, continual, and reinforcement learning.

Stochastic Gradient Langevin Dynamics, the "unadjusted Langevin algorithm", and Adaptive Langevin Dynamics (also known as Stochastic Gradient Nos\'{e}-Hoover dynamics) are examples of existing thermodynamic parameterization methods in use for machine learning, but these can be substantially improved. We find that by partitioning the parameters based on natural layer structure we obtain schemes with rapid convergence for data sets with complicated loss landscapes. We describe easy-to-implement hybrid partitioned numerical algorithms, based on discretized stochastic differential equations, which are adapted to feed-forward neural networks, including LaLa (a multi-layer Langevin algorithm), AdLaLa (combining the adaptive Langevin and Langevin algorithms) and LOL (combining Langevin and Overdamped Langevin); we examine the convergence of these methods using numerical studies and compare their performance among themselves and in relation to standard alternatives such as stochastic gradient descent and ADAM. We present evidence that thermodynamic parameterization methods can be (i) faster, (ii) more accurate, and (iii) more robust than standard algorithms incorporated into machine learning frameworks, in particular for data sets with complicated loss landscapes. Moreover, we show in numerical studies that methods based on sampling excite many degrees of freedom. The equipartition property, which is a consequence of their ergodicity, means that these methods keep in play an ensemble of low-loss states during the training process. By drawing parameter states from a sufficiently rich distribution of nearby candidate states, we show that the thermodynamic schemes produce smoother classifiers, improve generalization and reduce overfitting compared to traditional optimizers.

In this work we investigate to which extent one can recover class probabilities within the empirical risk minimization (ERM) paradigm. The main aim of our paper is to extend existing results and emphasize the tight relations between empirical risk minimization and class probability estimation. Based on existing literature on excess risk bounds and proper scoring rules, we derive a class probability estimator based on empirical risk minimization. We then derive fairly general conditions under which this estimator will converge, in the L1-norm and in probability, to the true class probabilities. Our main contribution is to present a way to derive finite sample L1-convergence rates of this estimator for different surrogate loss functions. We also study in detail which commonly used loss functions are suitable for this estimation problem and finally discuss the setting of model-misspecification as well as a possible extension to asymmetric loss functions.