Publications
Below is a list of my publications together with a link to a more in-depth description.
New publications are gradually added as they become available.
2024
Evtonomy: A Personalised Route Planner For Electric Vehicles
A. Augustin, E. Shafipour, S. Stein,
Proceedings of the 23rd International Conference on Autonomous Agents and Multi-Agent Systems (AAMAS),
2024.
Abstract: With the continuing growth of the elect… Abstract: With the continuing growth of the electric vehicle (EV) market, planning long road trips should be a seamless and hassle-free experience for EV owners. Dedicated EV route planning apps have emerged recently as indispensable assistants providing essential mapping and data services. However, EV owners still face a number of challenges when planning their routes to prevent unnecessary delays or expenses. These challenges are not yet fully addressed with current EV planning apps. This paper introduces EVtonomy, an app that assigns an intelligent agent to each driver capable of planning personalised journeys. Specifically, the agent provides routes and charging stop recommendations aligned with the EV owner's individual preferences in terms of trip duration, including both driving time and the time spent charging the car, along with the total charging costs. Read more Read less
#research
#publications
#Electric Vehicles
#Online Route Planning
2020
The Impact Of Therapeutic Inertia On Long-Term Blood Pressure Control: A Monte Carlo Simulation Study
A. Augustin, F. Shankar, L. Coutts, C. N. Floyd, A. S. Wierzbicki, L. Zanisi, P. J. Chowienczyk,
Journal of Hypertension,
2020.
Abstract: Antihypertensive drug treatment is base… Abstract: Antihypertensive drug treatment is based on stepped titration in response to elevated blood pressure (BP) measurements. However, measurements do not necessarily represent an individual's true BP (due to random error and biological variability), and medications are not always increased when measurements are above target (therapeutic inertia). We developed a Monte Carlo model with a 10-year horizon to investigate how measurement error impacted systolic BP (SBP) control in the presence of therapeutic inertia. When SBP measurements were in the range 140 to 159 mmHg, the probability of escalating treatment was determined by a Bernoulli probability mass function parameterized by weighting functions exploring distinct inertia profiles. Simulating inertia with the weighting function that approximated to clinical practice resulted in ≈50% of individuals failing to achieve their SBP target within the 10-year time horizon. An inverse relationship was observed between measurement error and SBP control. This suggests that the value of accurate SBP measurement is only realized if it changes the underlying probability of inertia—that is, patients/clinicians believe a measurement to be accurate and so are more likely to act upon it. Removal of inertia during treatment initiation (ie, stepped titration until SBP measurement was below target) improved true SBP control for all simulations. Our simulations show that the impact of therapeutic inertia during treatment initiation persists during long-term follow-up. Strategies to remove therapeutic inertia during treatment initiation (ie, dual antihypertensive therapy) are likely to improve long-term BP control irrespective of BP measurement technique. Read more Read less
#statistics
#healthcare
#research
#publications
2020
Open-Domain Topic Identification Of Out-Of-Domain Utterances Using Wikipedia
A. Augustin, A. Papangelis, M. Kotti, P. Vougiouklis, J. Hare, N. Braunschweiler,
Thirty-fourth Conference on Neural Information Processing Systems (NeurIPS), Workshop on Human in the Loop Dialogue Systems (HLDS),
2020.
Abstract: Users of spoken dialogue systems (SDS) … Abstract: Users of spoken dialogue systems (SDS) expect high quality interactions across a wide range of diverse topics. However, the implementation of SDS capable of responding to every conceivable user utterance in an informative way is a challenging problem. Multi-domain SDS must necessarily identify and deal with out-of-domain (OOD) utterances to generate appropriate responses as users do not always know in advance what domains the SDS can handle. To address this problem, we extend the current state-of-the-art in multi-domain SDS by estimating the topic of OOD utterances using external knowledge representation from Wikipedia. Experimental results on real human-to-human dialogues showed that our approach does not degrade domain prediction performance when compared to the base model. But more significantly, our joint training achieves more accurate predictions of the nearest Wikipedia article by up to about 30% when compared to the benchmarks. Read more Read less
#machinelearning
#research
#publications
2020
Learning And Evaluation Of Topics Via Distributional Semantics
A. Augustin,
University of Southampton Institutional Repository,
2020.
Abstract: Written language is a means of communic… Abstract: Written language is a means of communication. It not only shapes our thoughts, written language also helps us communicate information. As the amount of digital text available keeps growing, it becomes increasingly difficult to locate and keep track of specific information of interest. This observation has fuelled the search for sophisticated representations of written text, and methods for learning meaning. In particular, topic identification has grown in importance in recent years as an approach to summarise, organise and understand text. Underpinning modern topic identification methods is the framework of distributional semantics which is based on the assumption that meaning is associated with use, and in particular, meaning can be learned by examining the contexts in which words occurs. Motivated by this, we look in this thesis at the broad field of topic identification in text learned via state-of-the-art distributional semantics models. As such, we provide new answers to the complex question of how meaning is used to derive abstract concepts like topics, and how non-expert humans evaluate such abstract concept generated from artificial processes. In more detail, we address three key problems. We first tackle the problem of evaluating the output of topic models (a particular kind of topic identification method) on large text corpora by leveraging non-expert annotators to assess the relevance of topics to a set of documents. Second, we develop a new method to assist in the interpretation of topics by providing additional context. In particular, our solution learns topics as collections of sentences extracted from large corpus of unstructured documents. Finally, we identify and track the topic of text collected over time. In particular, we look at text-based dialogues which often consists of short utterances covering a variety of topics. Read more Read less
#machinelearning
#research
#publications
2017
Bayesian Aggregation Of Categorical Distributions With Applications In Crowdsourcing
A. Augustin, M. Venanzi, A. Rogers, N.R. Jennings,
Proceedings of the 26th International Joint Conference on Artificial Intelligence (IJCAI),
2017.
Abstract: A key problem in crowdsourcing is the a… Abstract: A key problem in crowdsourcing is the aggregation of judgments of proportions. For example, workers might be presented with a news article or an image, and be asked to identify the proportion of each topic, sentiment, object, or colour present in it. These varying judgments then need to be aggregated to form a consensus view of the document's or image's contents. Often, however, these judgments are skewed by workers who provide judgments randomly. Such spammers make the cost of acquiring judgments more expensive and degrade the accuracy of the aggregation. For such cases, we provide a new Bayesian framework for aggregating these responses (expressed in the form of categorical distributions) that for the first time accounts for spammers. We elicit 796 judgments about proportions of objects and coloursin images. Experimental results show comparable aggregation accuracy when 60% of the workers are spammers, as other state of the art approaches do when there are no spammers. Read more Read less
#machinelearning
#research
#publications
2011
A Parametric Approach To Bayesian Drift Estimation Of Discretely Observed Price Diffusion Processes
A. Augustin,
Imperial College London,
2011.
Abstract: We suggest an approach to parametric es… Abstract: We suggest an approach to parametric estimation of the drift function mu(t) in a one-dimensional Ito stochastic differential equation representing a stock price with a variable diffusion parameter sigma such that dSt = mu(t)Stdt + sigmaStdWt from a portfolio consisting of the stock, a call option and a forward contract. The parametric charactrisation of mu(t) is derived from non-arbitrage arguments used to price derivatives. The stock price process is expressed in terms of discretely observed derivative prices and the unobservable discrete market risk premium, the later modeled using the Ornstein-Uhlenbeck process. To address the time variation of volatility sigma, the Black-Scholes implied volatility of the option is used. The resulting system of prices and market risk premium dynamics is then be cast into the state space framework from which the time varying conditional distribution of the market risk premium can be estimated using the Kalman filter. Numerical analysis are carried out using simulated market prices. Read more Read less
#machinelearning
#finance
#research
#publications
Preprint
Quantitative Strategies For The Personalisation Of Antihypertensive Therapy
A. Augustin, F. Shankar, D. Burns, C. M. Baker-Smith, L. Coutts, Phil J. Chowienczykb, C. N. Floyd,
preprint,
.
Abstract: We propose a computationally efficient … Abstract: We propose a computationally efficient methodology to determine the probability of pre- and post-medication SBP to be below the recommended treatment threshold for an individual after one or more medications intake from a single SBP measurement. We build upon a prior probabilistic framework that model individuals’ true SBP as a latent variable and makes predictions of pre- and post-medication SBP via Monte Carlo (MC) sampling. We extend this approach by introducing a computationally efficient formulation based on numerical recursive convolutions which also takes into account prior history of measured SBP across multiple visits via a Bayesian update mechanism. Our methodology produces results indistinguishable with state-of-art MC simulations irrespective of the experimental setting (i.e. measured SBP, standard deviation of the drug and measurement) while being less resource intensive. Moreover, results showed that the inclusion of historical data in the analysis leads to more accurate prediction of future true SBP. Read more Read less
#statistics
#healthcare
#research
#publications