Santu Rana

Recognising faces in unseen modes: A tensor based approach

This paper addresses the limitation of current multilinear techniques (multilinear PCA, multilinear ICA) when applied to face recognition for handling faces in unseen illumination and viewpoints. We propose a new recognition method, exploiting the interaction of all the subspaces resulting from multilinear decomposition (for both multilinear PCA and ICA), to produce a new basis called multilinear-eigenmodes. This basis offers the flexibility to handle face images at unseen illumination or viewpoints. Experiments on benchmarked datasets yield superior performance in terms of both accuracy and computational cost.

Predicting unplanned readmission after myocardial infarction from routinely collected administrative hospital data

OBJECTIVE Readmission rates are high following acute myocardial infarction (AMI), but risk stratification has proved difficult because known risk factors are only weakly predictive. In the present study, we applied hospital data to identify the risk of unplanned admission following AMI hospitalisations. METHODS The study included 1660 consecutive AMI admissions. Predictive models were derived from 1107 randomly selected records and tested on the remaining 553 records. The electronic medical record (EMR) model was compared with a seven-factor predictive score known as the HOSPITAL score and a model derived from Elixhauser comorbidities. All models were evaluated for the ability to identify patients at high risk of 30-day ischaemic heart disease readmission and those at risk of all-cause readmission within 12 months following the initial AMI hospitalisation. RESULTS The EMR model has higher discrimination than other models in predicting ischaemic heart disease readmissions (area under the curve (AUC) 0.78; 95% confidence interval (CI) 0.71-0.85 for 30-day readmission). The positive predictive value was significantly higher with the EMR model, which identifies cohorts that were up to threefold more likely to be readmitted. Factors associated with readmission included emergency department attendances, cardiac diagnoses and procedures, renal impairment and electrolyte disturbances. The EMR model also performed better than other models (AUC 0.72; 95% CI 0.66-0.78), and with greater positive predictive value, in identifying 12-month risk of all-cause readmission. CONCLUSIONS Routine hospital data can help identify patients at high risk of readmission following AMI. This could lead to decreased readmission rates by identifying patients suitable for targeted clinical interventions.

Differentially Private Random Forest with High Utility

Privacy-preserving data mining has become an active focus of the research community in the domains where data are sensitive and personal in nature. For example, highly sensitive digital repositories of medical or financial records offer enormous values for risk prediction and decision making. However, prediction models derived from such repositories should maintain strict privacy of individuals. We propose a novel random forest algorithm under the framework of differential privacy. Unlike previous works that strictly follow differential privacy and keep the complete data distribution approximately invariant to change in one data instance, we only keep the necessary statistics (e.g. variance of the estimate) invariant. This relaxation results in significantly higher utility. To realize our approach, we propose a novel differentially private decision tree induction algorithm and use them to create an ensemble of decision trees. We also propose feasible adversary models to infer about the attribute and class label of unknown data in presence of the knowledge of all other data. Under these adversary models, we derive bounds on the maximum number of trees that are allowed in the ensemble while maintaining privacy. We focus on binary classification problem and demonstrate our approach on four real-world datasets. Compared to the existing privacy preserving approaches we achieve significantly higher utility.

Web search activity data accurately predict population chronic disease risk in the USA

Background The WHO framework for non-communicable disease (NCD) describes risks and outcomes comprising the majority of the global burden of disease. These factors are complex and interact at biological, behavioural, environmental and policy levels presenting challenges for population monitoring and intervention evaluation. This paper explores the utility of machine learning methods applied to population-level web search activity behaviour as a proxy for chronic disease risk factors. Methods Web activity output for each element of the WHOs Causes of NCD framework was used as a basis for identifying relevant web search activity from 2004 to 2013 for the USA. Multiple linear regression models with regularisation were used to generate predictive algorithms, mapping web search activity to Centers for Disease Control and Prevention (CDC) measured risk factor/disease prevalence. Predictions for subsequent target years not included in the model derivation were tested against CDC data from population surveys using Pearson correlation and Spearmans r. Results For 2011 and 2012, predicted prevalence was very strongly correlated with measured risk data ranging from fruits and vegetables consumed (r=0.81; 95% CI 0.68 to 0.89) to alcohol consumption (r=0.96; 95% CI 0.93 to 0.98). Mean difference between predicted and measured differences by State ranged from 0.03 to 2.16. Spearmans r for state-wise predicted versus measured prevalence varied from 0.82 to 0.93. Conclusions The high predictive validity of web search activity for NCD risk has potential to provide real-time information on population risk during policy implementation and other population-level NCD prevention efforts.

Large-scale statistical modeling of motion patterns: a Bayesian nonparametric approach

We propose a novel framework for large-scale scene understanding in static camera surveillance. Our techniques combine fast rank-1 constrained robust PCA to compute the foreground, with non-parametric Bayesian models for inference. Clusters are extracted in foreground patterns using a joint multinomial+Gaussian Dirichlet process model (DPM). Since the multinomial distribution is normalized, the Gaussian mixture distinguishes between similar spatial patterns but different activity levels (eg. car vs bike). We propose a modification of the decayed MCMC technique for incremental inference, providing the ability to discover theoretically unlimited patterns in unbounded video streams. A promising by-product of our framework is online, abnormal activity detection. A benchmark video and two surveillance videos, with the longest being 140 hours long are used in our experiments. The patterns discovered are as informative as existing scene understanding algorithms. However, unlike existing work, we achieve near real-time execution and encouraging performance in abnormal activity detection.

Flexible Transfer Learning Framework for Bayesian Optimisation

Bayesian optimisation is an efficient technique to optimise functions that are expensive to compute. In this paper, we propose a novel framework to transfer knowledge from a completed source optimisation task to a new target task in order to overcome the cold start problem. We model source data as noisy observations of the target function. The level of noise is computed from the data in a Bayesian setting. This enables flexible knowledge transfer across tasks with differing relatedness, addressing a limitation of the existing methods. We evaluate on the task of tuning hyperparameters of two machine learning algorithms. Treating a fraction of the whole training data as source and the whole as the target task, we show that our method finds the best hyperparameters in the least amount of time compared to both the state-of-art and no transfer method.

Rapid Bayesian optimisation for synthesis of short polymer fiber materials.

The discovery of processes for the synthesis of new materials involves many decisions about process design, operation, and material properties. Experimentation is crucial but as complexity increases, exploration of variables can become impractical using traditional combinatorial approaches. We describe an iterative method which uses machine learning to optimise process development, incorporating multiple qualitative and quantitative objectives. We demonstrate the method with a novel fluid processing platform for synthesis of short polymer fibers, and show how the synthesis process can be efficiently directed to achieve material and process objectives.

Budgeted Batch Bayesian Optimization

Parameter settings profoundly impact the performance of machine learning algorithms and laboratory experiments. The classical trial-error methods are exponentially expensive in large parameter spaces, and Bayesian optimization (BO) offers an elegant alternative for global optimization of black box functions. In situations where the functions can be evaluated at multiple points simultaneously, batch Bayesian optimization is used. Current batch BO approaches are restrictive in fixing the number of evaluations per batch, and this can be wasteful when the number of specified evaluations is larger than the number of real maxima in the underlying acquisition function. We present the budgeted batch Bayesian optimization (B3O) for hyper-parameter tuning and experimental design - we identify the appropriate batch size for each iteration in an elegant way. In particular, we use the infinite Gaussian mixture model (IGMM) for automatically identifying the number of peaks in the underlying acquisition functions. We solve the intractability of estimating the IGMM directly from the acquisition function by formulating the batch generalized slice sampling to efficiently draw samples from the acquisition function. We perform extensive experiments for benchmark functions and two real world applications - machine learning hyper-parameter tuning and experimental design for alloy hardening. We show empirically that the proposed B3O outperforms the existing fixed batch BO approaches in finding the optimum whilst requiring a fewer number of evaluations, thus saving cost and time.

Is demography destiny? Application of machine learning techniques to accurately predict population health outcomes from a minimal demographic dataset.

For years, we have relied on population surveys to keep track of regional public health statistics, including the prevalence of non-communicable diseases. Because of the cost and limitations of such surveys, we often do not have the up-to-date data on health outcomes of a region. In this paper, we examined the feasibility of inferring regional health outcomes from socio-demographic data that are widely available and timely updated through national censuses and community surveys. Using data for 50 American states (excluding Washington DC) from 2007 to 2012, we constructed a machine-learning model to predict the prevalence of six non-communicable disease (NCD) outcomes (four NCDs and two major clinical risk factors), based on population socio-demographic characteristics from the American Community Survey. We found that regional prevalence estimates for non-communicable diseases can be reasonably predicted. The predictions were highly correlated with the observed data, in both the states included in the derivation model (median correlation 0.88) and those excluded from the development for use as a completely separated validation sample (median correlation 0.85), demonstrating that the model had sufficient external validity to make good predictions, based on demographics alone, for areas not included in the model development. This highlights both the utility of this sophisticated approach to model development, and the vital importance of simple socio-demographic characteristics as both indicators and determinants of chronic disease.

Exploiting side information in distance dependent Chinese restaurant processes for data clustering

Multimedia contents often possess weakly annotated data such as tags, links and interactions. The weakly annotated data is called side information. It is the auxiliary information of data and provides hints for exploring the link structure of data. Most clustering algorithms utilize pure data for clustering. A model that combines pure data and side information, such as images and tags, documents and keywords, can perform better at understanding the underlying structure of data. We demonstrate how to incorporate different types of side information into a recently proposed Bayesian nonparametric model, the distance dependent Chinese restaurant process (DD-CRP). Our algorithm embeds the affinity of this information into the decay function of the DD-CRP when side information is in the form of subsets of discrete labels. It is flexible to measure distance based on arbitrary side information instead of only the spatial layout or time stamp of observations. At the same time, for noisy and incomplete side information, we set the decay function so that the DD-CRP reduces to the traditional Chinese restaurant process, thus not inducing side effects of noisy and incomplete side information. Experimental evaluations on two real-world datasets NUS WIDE and 20 Newsgroups show exploiting side information in DD-CRP significantly improves the clustering performance.