Searcharxiv⌕ Search

arXiv subjects

Chris Holmes

Publications and source records attributed to Chris Holmes.

At least 73 records · Page 4Linked to original sources
arXiv2020

Foundations of Bayesian Learning from Synthetic Data

There is significant growth and interest in the use of synthetic data as an enabler for machine learning in environments where the release of real data is restricted due to privacy or availability constraints. Despite a large number of methods for synthetic data generation, there are comparatively few results on the statistical properties of models learnt on synthetic data, and fewer still for situations where a researcher wishes to augment real data with another party's synthesised data. We use a Bayesian paradigm to characterise the updating of model parameters when learning in these settings, demonstrating that caution should be taken when applying conventional learning algorithms without appropriate consideration of the synthetic data generating process and learning task. Recent results from general Bayesian updating support a novel and robust approach to Bayesian synthetic-learning founded on decision theory that outperforms standard approaches across repeated experiments on supervised learning and inference problems.

cs.LG↗
arXiv2020

Inferring proximity from Bluetooth Low Energy RSSI with Unscented Kalman Smoothers

The Covid-19 pandemic has resulted in a variety of approaches for managing infection outbreaks in international populations. One example is mobile phone applications, which attempt to alert infected individuals and their contacts by automatically inferring two key components of infection risk: the proximity to an individual who may be infected, and the duration of proximity. The former component, proximity, relies on Bluetooth Low Energy (BLE) Received Signal Strength Indicator(RSSI) as a distance sensor, and this has been shown to be problematic; not least because of unpredictable variations caused by different device types, device location on-body, device orientation, the local environment and the general noise associated with radio frequency propagation. In this paper, we present an approach that infers posterior probabilities over distance given sequences of RSSI values. Using a single-dimensional Unscented Kalman Smoother (UKS) for non-linear state space modelling, we outline several Gaussian process observation transforms, including: a generative model that directly captures sources of variation; and a discriminative model that learns a suitable observation function from training data using both distance and infection risk as optimisation objective functions. Our results show that good risk prediction can be achieved in $\mathcal{O}(n)$ time on real-world data sets, with the UKS outperforming more traditional classification methods learned from the same training data.

eess.SP↗
arXiv2019

Disentangling to Cluster: Gaussian Mixture Variational Ladder Autoencoders

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.

cs.LG↗
arXiv2019

Regularising Deep Networks with Deep Generative Models

We develop a new method for regularising neural networks. We learn a probability distribution over the activations of all layers of the model and then insert imputed values into the network during training. We obtain a posterior for an arbitrary subset of activations conditioned on the remainder. This is a generalisation of data augmentation to the hidden layers of a network, and a form of data-aware dropout. We demonstrate that our training method leads to higher test accuracy and lower test-set cross-entropy for neural networks trained on CIFAR-10 and SVHN compared to standard regularisation baselines: our approach leads to networks with better calibrated uncertainty over the class posteriors all the while delivering greater test-set accuracy.

cs.LG↗
arXiv2019

On the marginal likelihood and cross-validation

In Bayesian statistics, the marginal likelihood, also known as the evidence, is used to evaluate model fit as it quantifies the joint probability of the data under the prior. In contrast, non-Bayesian models are typically compared using cross-validation on held-out data, either through $k$-fold partitioning or leave-$p$-out subsampling. We show that the marginal likelihood is formally equivalent to exhaustive leave-$p$-out cross-validation averaged over all values of $p$ and all held-out test sets when using the log posterior predictive probability as the scoring rule. Moreover, the log posterior predictive is the only coherent scoring rule under data exchangeability. This offers new insight into the marginal likelihood and cross-validation and highlights the potential sensitivity of the marginal likelihood to the choice of the prior. We suggest an alternative approach using cumulative cross-validation following a preparatory training phase. Our work has connections to prequential analysis and intrinsic Bayes factors but is motivated through a different course.

stat.ME↗
arXiv2019

Scalable Nonparametric Sampling from Multimodal Posteriors with the Posterior Bootstrap

Increasingly complex datasets pose a number of challenges for Bayesian inference. Conventional posterior sampling based on Markov chain Monte Carlo can be too computationally intensive, is serial in nature and mixes poorly between posterior modes. Further, all models are misspecified, which brings into question the validity of the conventional Bayesian update. We present a scalable Bayesian nonparametric learning routine that enables posterior sampling through the optimization of suitably randomized objective functions. A Dirichlet process prior on the unknown data distribution accounts for model misspecification, and admits an embarrassingly parallel posterior bootstrap algorithm that generates independent and exact samples from the nonparametric posterior distribution. Our method is particularly adept at sampling from multimodal posterior distributions via a random restart mechanism. We demonstrate our method on Gaussian mixture model and sparse logistic regression examples.

stat.ML↗
arXiv2019

A Framework for Adaptive MCMC Targeting Multimodal Distributions

We propose a new Monte Carlo method for sampling from multimodal distributions. The idea of this technique is based on splitting the task into two: finding the modes of a target distribution $π$ and sampling, given the knowledge of the locations of the modes. The sampling algorithm relies on steps of two types: local ones, preserving the mode; and jumps to regions associated with different modes. Besides, the method learns the optimal parameters of the algorithm while it runs, without requiring user intervention. Our technique should be considered as a flexible framework, in which the design of moves can follow various strategies known from the broad MCMC literature. In order to design an adaptive scheme that facilitates both local and jump moves, we introduce an auxiliary variable representing each mode and we define a new target distribution $\tildeπ$ on an augmented state space $\mathcal{X}~\times~\mathcal{I}$, where $\mathcal{X}$ is the original state space of $π$ and $\mathcal{I}$ is the set of the modes. As the algorithm runs and updates its parameters, the target distribution $\tildeπ$ also keeps being modified. This motivates a new class of algorithms, Auxiliary Variable Adaptive MCMC. We prove general ergodic results for the whole class before specialising to the case of our algorithm.

stat.CO↗
arXiv2018

Machine learning and AI research for Patient Benefit: 20 Critical Questions on Transparency, Replicability, Ethics and Effectiveness

Machine learning (ML), artificial intelligence (AI) and other modern statistical methods are providing new opportunities to operationalize previously untapped and rapidly growing sources of data for patient benefit. Whilst there is a lot of promising research currently being undertaken, the literature as a whole lacks: transparency; clear reporting to facilitate replicability; exploration for potential ethical concerns; and, clear demonstrations of effectiveness. There are many reasons for why these issues exist, but one of the most important that we provide a preliminary solution for here is the current lack of ML/AI- specific best practice guidance. Although there is no consensus on what best practice looks in this field, we believe that interdisciplinary groups pursuing research and impact projects in the ML/AI for health domain would benefit from answering a series of questions based on the important issues that exist when undertaking work of this nature. Here we present 20 questions that span the entire project life cycle, from inception, data analysis, and model evaluation, to implementation, as a means to facilitate project planning and post-hoc (structured) independent evaluation. By beginning to answer these questions in different settings, we can start to understand what constitutes a good answer, and we expect that the resulting discussion will be central to developing an international consensus framework for transparent, replicable, ethical and effective research in artificial intelligence (AI-TREE) for health.

cs.CY↗
arXiv2018

Semi-unsupervised Learning of Human Activity using Deep Generative Models

We introduce 'semi-unsupervised learning', a problem regime related to transfer learning and zero-shot learning where, in the training data, some classes are sparsely labelled and others entirely unlabelled. Models able to learn from training data of this type are potentially of great use as many real-world datasets are like this. Here we demonstrate a new deep generative model for classification in this regime. Our model, a Gaussian mixture deep generative model, demonstrates superior semi-unsupervised classification performance on MNIST to model M2 from Kingma and Welling (2014). We apply the model to human accelerometer data, performing activity classification and structure discovery on windows of time series data.

stat.ML↗
arXiv2018

General Bayesian Updating and the Loss-Likelihood Bootstrap

In this paper we revisit the weighted likelihood bootstrap, a method that generates samples from an approximate Bayesian posterior of a parametric model. We show that the same method can be derived, without approximation, under a Bayesian nonparametric model with the parameter of interest defined as minimising an expected negative log-likelihood under an unknown sampling distribution. This interpretation enables us to extend the weighted likelihood bootstrap to posterior sampling for parameters minimizing an expected loss. We call this method the loss-likelihood bootstrap. We make a connection between this and general Bayesian updating, which is a way of updating prior belief distributions without needing to construct a global probability model, yet requires the calibration of two forms of loss function. The loss-likelihood bootstrap is used to calibrate the general Bayesian posterior by matching asymptotic Fisher information. We demonstrate the methodology on a number of examples.

stat.ME↗
arXiv2018

Principles of Bayesian Inference using General Divergence Criteria

When it is acknowledged that all candidate parameterised statistical models are misspecified relative to the data generating process, the decision maker (DM) must currently concern themselves with inference for the parameter value minimising the KL-divergence between the model and the process (Walker, 2013). However, it has long been known that minimising the KL-divergence places a large weight on correctly capturing the tails of the sample distribution. As a result the DM is required to worry about the robustness of their model to tail misspecifications if they want to conduct principled inference. In this paper we alleviate these concerns for the DM. We advance recent methodological developments in general Bayesian updating (Bissiri, Holmes and Walker, 2016) to propose a statistically well principled Bayesian updating of beliefs targeting the minimisation of more general divergence criteria. We improve both the motivation and the statistical foundations of existing Bayesian minimum divergence estimation (Hooker and Vidyashankar, 2014; Ghosh and Basu, 2016), allowing the well principled Bayesian to target predictions from the model that are close to the genuine model in terms of some alternative divergence measure to the KL-divergence. Our principled formulation allows us to consider a broader range of divergences than have previously been considered. In fact we argue defining the divergence measure forms an important, subjective part of any statistical analysis, and aim to provide some decision theoretic rational for this selection. We illustrate how targeting alternative divergence measures can impact the conclusions of simple inference tasks, and discuss then how our methods might apply to more complicated, high dimensional models.

math.ST↗
arXiv2017

Multiscale Blind Source Separation

We provide a new methodology for statistical recovery of single linear mixtures of piecewise constant signals (sources) with unknown mixing weights and change points in a multiscale fashion. We show exact recovery within an $ε$-neighborhood of the mixture when the sources take only values in a known finite alphabet. Based on this we provide the SLAM (Separates Linear Alphabet Mixtures) estimators for the mixing weights and sources. For Gaussian error, we obtain uniform confidence sets and optimal rates (up to log-factors) for all quantities. SLAM is efficiently computed as a nonconvex optimization problem by a dynamic program tailored to the finite alphabet assumption. Its performance is investigated in a simulation study. Finally, it is applied to assign copy-number aberrations from genetic sequencing data to different clones and to estimate their proportions.

stat.ME↗
arXiv2017

Assigning a value to a power likelihood in a general Bayesian model

Bayesian approaches to data analysis and machine learning are widespread and popular as they provide intuitive yet rigorous axioms for learning from data; see Bernardo and Smith (2004) and Bishop (2006). However, this rigour comes with a caveat that the Bayesian model is a precise reflection of Nature. There has been a recent trend to address potential model misspecification by raising the likelihood function to a power, primarily for robustness reasons, though not exclusively. In this paper we provide a coherent specification of the power parameter once the Bayesian model has been specified in the absence of a perfect model.

stat.ME↗
arXiv2016

A Bayesian nonparametric approach to testing for dependence between random variables

Nonparametric and nonlinear measures of statistical dependence between pairs of random variables are important tools in modern data analysis. In particular the emergence of large data sets can now support the relaxation of linearity assumptions implicit in traditional association scores such as correlation. Here we describe a Bayesian nonparametric procedure that leads to a tractable, explicit and analytic quantification of the relative evidence for dependence vs independence. Our approach uses Polya tree priors on the space of probability measures which can then be embedded within a decision theoretic test for dependence. Polya tree priors can accommodate known uncertainty in the form of the underlying sampling distribution and provides an explicit posterior probability measure of both dependence and independence. Well known advantages of having an explicit probability measure include: easy comparison of evidence across different studies; encoding prior information; quantifying changes in dependence across different experimental conditions, and; the integration of results within formal decision analysis.

stat.ME↗
arXiv2016

A General Framework for Updating Belief Distributions

We propose a framework for general Bayesian inference. We argue that a valid update of a prior belief distribution to a posterior can be made for parameters which are connected to observations through a loss function rather than the traditional likelihood function, which is recovered under the special case of using self information loss. Modern application areas make it is increasingly challenging for Bayesians to attempt to model the true data generating mechanism. Moreover, when the object of interest is low dimensional, such as a mean or median, it is cumbersome to have to achieve this via a complete model for the whole data distribution. More importantly, there are settings where the parameter of interest does not directly index a family of density functions and thus the Bayesian approach to learning about such parameters is currently regarded as problematic. Our proposed framework uses loss-functions to connect information in the data to functionals of interest. The updating of beliefs then follows from a decision theoretic approach involving cumulative loss functions. Importantly, the procedure coincides with Bayesian updating when a true likelihood is known, yet provides coherent subjective inference in much more general settings. Connections to other inference frameworks are highlighted.

math.ST↗
arXiv2015

Scalable Bayesian nonparametric regression via a Plackett-Luce model for conditional ranks

We present a novel Bayesian nonparametric regression model for covariates X and continuous, real response variable Y. The model is parametrized in terms of marginal distributions for Y and X and a regression function which tunes the stochastic ordering of the conditional distributions F(y|x). By adopting an approximate composite likelihood approach, we show that the resulting posterior inference can be decoupled for the separate components of the model. This procedure can scale to very large datasets and allows for the use of standard, existing, software from Bayesian nonparametric density estimation and Plackett-Luce ranking estimation to be applied. As an illustration, we show an application of our approach to a US Census dataset, with over 1,300,000 data points and more than 100 covariates.

stat.ME↗
arXiv2015

On Markov chain Monte Carlo methods for tall data

Markov chain Monte Carlo methods are often deemed too computationally intensive to be of any practical use for big data applications, and in particular for inference on datasets containing a large number $n$ of individual data points, also known as tall datasets. In scenarios where data are assumed independent, various approaches to scale up the Metropolis-Hastings algorithm in a Bayesian inference context have been recently proposed in machine learning and computational statistics. These approaches can be grouped into two categories: divide-and-conquer approaches and, subsampling-based algorithms. The aims of this article are as follows. First, we present a comprehensive review of the existing literature, commenting on the underlying assumptions and theoretical guarantees of each method. Second, by leveraging our understanding of these limitations, we propose an original subsampling-based approach which samples from a distribution provably close to the posterior distribution of interest, yet can require less than $O(n)$ data point likelihood evaluations at each iteration for certain statistical models in favourable scenarios. Finally, we have only been able so far to propose subsampling-based methods which display good performance in scenarios where the Bernstein-von Mises approximation of the target posterior distribution is excellent. It remains an open challenge to develop such methods in scenarios where the Bernstein-von Mises approximation is poor.

stat.ME↗