With the overwhelming online products available in recent years, there is an increasing need to filter and deliver relevant personalized advice for users. Recommender systems solve this problem by modeling and predicting individual preferences for a great variety of items such as movies, books or research articles. In this chapter, we explore rigorous network-based models that outperform leading approaches for recommendation. The network models we consider are based on the explicit assumption that there are groups of individuals and of items, and that the preferences of an individual for an item are determined only by their group memberships. The accurate prediction of individual user preferences over items can be accomplished by different methodologies, such as Monte Carlo sampling or Expectation-Maximization methods, the latter resulting in a scalable algorithm which is suitable for large datasets.

Supervised learning requires the specification of a loss function to minimise. While the theory of admissible losses from both a computational and statistical perspective is well-developed, these offer a panoply of different choices. In practice, this choice is typically made in an \emph{ad hoc} manner. In hopes of making this procedure more principled, the problem of \emph{learning the loss function} for a downstream task (e.g., classification) has garnered recent interest. However, works in this area have been generally empirical in nature. In this paper, we revisit the {\sc SLIsotron} algorithm of Kakade et al. (2011) through a novel lens, derive a generalisation based on Bregman divergences, and show how it provides a principled procedure for learning the loss. In detail, we cast {\sc SLIsotron} as learning a loss from a family of composite square losses. By interpreting this through the lens of \emph{proper losses}, we derive a generalisation of {\sc SLIsotron} based on Bregman divergences. The resulting {\sc BregmanTron} algorithm jointly learns the loss along with the classifier. It comes equipped with a simple guarantee of convergence for the loss it learns, and its set of possible outputs comes with a guarantee of agnostic approximability of Bayes rule. Experiments indicate that the {\sc BregmanTron} substantially outperforms the {\sc SLIsotron}, and that the loss it learns can be minimized by other algorithms for different tasks, thereby opening the interesting problem of \textit{loss transfer} between domains.

Tree-based algorithms for supervised learning, such as Classification and Regression Trees (CART) (Breiman et al., 1984), random forests (Breiman, 1996, 2001), adaBoost (Freund and Schapire, 1997), and gradient boosting (Breiman, 1997; Friedman, 2001, 2002), are widely used for applied supervised learning. As a whole, these methods are popular in applied settings due to their speed and accuracy in mean estimation and out-of-sample prediction tasks. One limitation of such methods is their well-known sensitivity to tuning parameters, which require costly cross-validation to optimize. Bayesian additive regression trees (BART) (Chipman et al., 2007, 2010) is a popular model-based alternative that is often more accurate than other treebased methods; specifically, BART boasts valuable robustness to the choice of tuning-parameters. However, relative to random forests and boosting, BART's wider adoption has been slowed by its more severe computational demands, owing to its reliance on a random walk Metropolis-Hastings Markov chain Monte Carlo (MCMC) algorithm. Despite this limitation, BART has inspired a considerable body of research in recent years.

Bayesian optimization is an effective method for finding extrema of a black-box function. We propose a new type of Bayesian optimization for learning user preferences in high-dimensional spaces. The central assumption is that the underlying objective function cannot be evaluated directly, but instead a minimizer along a projection can be queried, which we call a projective preferential query. The form of the query allows for feedback that is natural for a human to give, and which enables interaction. This is demonstrated in a user experiment in which the user feedback comes in the form of optimal position and orientation of a molecule adsorbing to a surface. We demonstrate that our framework is able to find a global minimum of a high-dimensional black-box function, which is an infeasible task for existing preferential Bayesian optimization frameworks that are based on pairwise comparisons.

Bayesian reinforcement learning (BRL) offers a decision-theoretic solution to the problem of reinforcement learning. However, typical model-based BRL algorithms have focused either on ma intaining a posterior distribution on models or value functions and combining this with approx imate dynamic programming or tree search. This paper describes a novel backwards induction pri nciple for performing joint Bayesian estimation of models and value functions, from which many new BRL algorithms can be obtained. We demonstrate this idea with algorithms and experiments in discrete state spaces.

Machine Learning Lecture at CMU by Ke Li, Ph.D. Candidate at the University of California, Berkeley Lecturer: Ke Li Carnegie Mellon University Abstract: In this talk, Li presents his team's work on overcoming two long-standing problems in machine learning and algorithms: 1. Mode collapse in generative adversarial nets (GANs) Generative adversarial nets (GANs) are perhaps the most popular class of generative models in use today. Unfortunately, they suffer from the well-documented problem of mode collapse, which the many successive variants of GANs have failed to overcome. I will illustrate why mode collapse happens fundamentally and show a simple way to overcome it, which is the basis of a new method known as Implicit Maximum Likelihood Estimation (IMLE). It turns out that this problem is not insurmountable - I will explain how the curse of dimensionality arises and show a simple way to overcome it, which gives rise to a new family of algorithms known as Dynamic Continuous Indexing (DCI). Bio: Ke Li is a recent Ph.D. graduate from UC Berkeley, where he was advised by Prof. Jitendra Malik, and will join Google as a Research Scientist and the Institute for Advanced Study (IAS) as a Member hosted by Prof. Sanjeev Arora.

Stochastic gradient Markov chain Monte Carlo (MCMC) algorithms have received much attention in Bayesian computing for big data problems, but they are only applicable to a small class of problems for which the parameter space has a fixed dimension and the log-posterior density is differentiable with respect to the parameters. This paper proposes an extended stochastic gradient MCMC lgoriathm which, by introducing appropriate latent variables, can be applied to more general large-scale Bayesian computing problems, such as those involving dimension jumping and missing data. Numerical studies show that the proposed algorithm is highly scalable and much more efficient than traditional MCMC algorithms. The proposed algorithms have much alleviated the pain of Bayesian methods in big data computing.

Variational Bayesian Inference is a popular methodology for approximating posterior distributions over Bayesian neural network weights. Recent work developing this class of methods has explored ever richer parameterizations of the approximate posterior in the hope of improving performance. In contrast, here we share a curious experimental finding that suggests instead restricting the variational distribution to a more compact parameterization. For a variety of deep Bayesian neural networks trained using Gaussian mean-field variational inference, we find that the posterior standard deviations consistently exhibit strong low-rank structure after convergence. This means that by decomposing these variational parameters into a low-rank factorization, we can make our variational approximation more compact without decreasing the models' performance. Furthermore, we find that such factorized parameterizations improve the signal-to-noise ratio of stochastic gradient estimates of the variational lower bound, resulting in faster convergence.

Mathematical models of system dynamics are a core technology in most model-based engineered systems acting and interacting with their environment. Examples include GPS, autonomous vehicles, passenger aircraft and robotics, to name just a few. The remarkable utility of mathematical models stems from the fact that, inter alia, they enable decision making based on prediction of system behaviour under new scenarios, accelerate the analysis and design processes, are fundamental to detecting faults or changes, and they are capable of handling uncertainty that is present in data, assumptions and algorithms. Motivated by the broad applicability and utility of modelling, the scientific community has devoted significant research attention towards learning dynamical models from data. Importantly, for dynamic systems, the sequence or ordering of the data must be maintained as future outcomes are deemed to be fundamentally related to the past. This is sometimes called sequence learning (Sun and Giles, 2001) or system identification (Ljung, 1999). In essence, these approaches search over a space of models and determine the model that best (in some sense) fits the data while maintaining the time ordering. The current paper is directed towards solving this important problem. To make these ideas more concrete, here we assume that data from the system of interest is available in the form of a data record y 1:T {y 1,...,y T }, where each measurementy k is potentially multidimensional and the number of available measurements is denoted as T 0. We further assume that the data may be adequately described as an instance from a joint distribution that is parametrized by an unknown vectorθ (called the parameter vector), that is (with abuse of notation)

Despite strong performance on a variety of tasks, neural sequence models trained with maximum likelihood have been shown to exhibit issues such as length bias and degenerate repetition. We study the related issue of receiving infinite-length sequences from a recurrent language model when using common decoding algorithms. To analyze this issue, we first define inconsistency of a decoding algorithm, meaning that the algorithm can yield an infinite-length sequence that has zero probability under the model. We prove that commonly used incomplete decoding algorithms - greedy search, beam search, top-k sampling, and nucleus sampling - are inconsistent, despite the fact that recurrent language models are trained to produce sequences of finite length. Based on these insights, we propose two remedies which address inconsistency: consistent variants of top-k and nucleus sampling, and a self-terminating recurrent language model. Empirical results show that inconsistency occurs in practice, and that the proposed methods prevent inconsistency.