# Doctoral Students

# Adilet Otemissov

## Oxford

## Bio

Adilet is a first year doctoral student from Oxford University. He graduated from Nazarbayev University with a BSc in Mathematics and then pursued a master’s degree at The University of Manchester. His research interests include combinatorial and non-convex optimisation.

## Research

Adilet’s PhD research will be dedicated to finding more efficient algorithms for the most difficult of optimisation problem classes, namely, that of global optimisation, where the best optima is desired amongst possibly-many local solutions in a highly nonlinear landscape, and often in large dimensions. The grand challenge is scalability of algorithms for global optimisation, as the state of the art software can generally solve global optimisation problems in the order of ten parameters. This project plans to explore modern techniques from compressed sensing, signal processing and machine learning – where techniques have been devised to find structure/important information in large data sets – to effectively reduce the dimension of global optimisation problems so that they can be solved efficiently. This is a little explored direction of research with great potential for impact in terms of algorithm development and transformative applications.

# Alex Bird

## Edinburgh

## Bio

Alex is a first year PhD student registered at Edinburgh, supervised by Chris Williams. His research interests lie primarily in graphical models and approximate inference.

## Research

Alex will be working on time series models for disambiguating artifactual signal(s). This will be a continuation of a project by Chris Williams and Konstantinos Georgatzis. The motivation stems from a collaboration with the Queen Elizabeth University Hospital in Glasgow, with the current ambition of predicting drug response in physiological monitoring data.

# Andreas Grammenos

## Cambridge

## Bio

Andreas was born in Athens, the capital of beautiful Greece in the late 80’s and formally obtained his Degree in Electronic & Computer Engineering from Technical University of Crete in 2015. He joined University of Cambridge as a first year PhD student in October 2016.

## Research

Andreas’ thesis revolves around the challenges involved when dealing with data at a massive scale which are primarily sourced from mobile devices. Finally, his general research interests include the broad field of Data Analytics which is an interesting intersection of computer science, systems programming and mathematics.

# Corinne Cath

## Oxford

## Bio

Corinne J.N. Cath worked at the intersection of Internet governance, tech-policy, and ethics for various years. She joined the Oxford Internet Institute in 2015 as an MSc student. Her MSc research at the OII focused on the (im)possibility of instantiating human rights in Internet standards. Before Oxford she was a policy advisor for a Democratic Congressman in Washington D.C. After graduating from the OII she worked for human rights NGO ARTICLE 19. Corinne has a BA in anthropology, MA in International relations, and MSc in Social Science of the Internet.

## Research

Corinne is currently a first-year DPhil (PhD) student at the University of Oxford. Her main research areas are the governance of the Internet’s technical infrastructure, and the related issues of Internet fragmentation, “responsibility-by-design”, and data ethics. Corinne’s other research interests include Artificial Intelligence, private-public cooperation in cybersecurity policy development, and the politics of Internet Governance organisations. She is working on these issues under the guidance of Professor Luciano Floridi.

# Daniel Wilson-Nunn

## Warwick

## Bio

Daniel has recently graduated from the University of Warwick with an integrated master’s degree in mathematics and statistics. He is now starting a PhD in statistics registered at the University of Warwick and based at the Turing. His areas of interest are rough path theory and machine learning, with particular application to sound.

## Research

Rough path theory is a method of analysis of SDEs driven by highly oscillatory processes, such as Brownian motion. Recent developments in the theory of rough paths have opened the doors for ideas and concepts developed for studying rough paths to be used in a wide variety of applied situations, including the recognition of Chinese handwriting and financial modelling. Combining the rich theory of rough paths to the rapidly growing field of machine learning and big data, the aim of Daniel’s project is to develop tools for sound analysis, recognition and compression.

# David Butler

## Edinburgh

## Bio

David is entering The Alan Turing Institute as a first year doctoral student. He is working with his supervisor, David Aspinall, registered with the University of Edinburgh. David’s research interests are in security and privacy. In particular he is working on the formalisation of the proof of security of cryptosystems and their protocols.

## Researcher

As cryptographic protocols become more complex it is becoming more difficult to prove that they are ‘secure’. That is to say, it is becoming very hard to ensure no secret information is leaked to an adversary that attacks the protocol. This increasing complexity has meant that proving security of these protocols by hand is becoming unfeasible; human mistakes and errors creep in. Consequently attempts are being made to formalise these security proofs on machine. David is currency working on formalising a general two party protocol scheme using the proof assistant, Isabelle, with particular interest in working with the Oblivious Transfer protocol.

# Edward Chuah

## Warwick

## Bio

Edward commenced his doctoral studies at the University of Warwick in October 2016. His general research interests are Reliability, fault/failure diagnosis, detection and prediction, anomaly detection and clustering, correlation and inference, large clusters and distributed systems, networking.

## Research

Recent research has demonstrated the value of combining message logs and resource use data for cluster system failure diagnosis. However, there is little work that combines analysis of patterns in message logs and resource use data to diagnose performance failures in cluster systems. Cluster systems have increasingly generated a massive amount of monitoring data which can be used to analyse these failures. Edward’s proposed research will study the nature and characteristics of system performance failures, use statistical and machine learning techniques to identify the causes of performance failures, develop new data-processing methodologies, implement tools for prototyping on production cluster systems.

# Helen Oliver

## Cambridge

## Bio

Helen Oliver is beginning her doctoral studies at The Alan Turing Institute after three years as a Research Associate at the University of Cambridge Computer Laboratory. She has been a researcher since 2008, and her research interests have converged at the intersection of wearable technology, narrative, privacy and innovation in the Internet of Things.

## Research

Helen’s project focuses on the research through design of wearable technology. As more users adopt wearable technology, more data of an ever more personal nature will be generated, raising urgent questions about how that data is handled and who benefits. At the same time, the HCI aspect of these devices is endangering mainstream acceptance: there is a lack of variety in the forms and functions of wearable devices, and a need to integrate wearables into IoT ecosystems. Helen’s research seeks concrete solutions for wearable technology, and the data it generates, to benefit the individual user and society as a whole.

# Javad Hosseini

## Edinburgh

## Bio

Javad started his PhD at University of Edinburgh in Autumn 2016. His research interests are in developing novel Machine Learning techniques to solve Natural Language Processing problems, especially in the area of semantics.

## Research

In particular, Javad’s interested in detecting the various statements that are made about the same identifiable entities of various types. He will use the datelines associated with the original stories to detect causal relations between event-types, as well as relations of paraphrase and entailment. The resulting semantics will be built into a semantic parser, which will be evaluated on standard datasets for question answering and text-entailment.

# Hilmi Majid

## Warwick

## Bio

Hilmi’s a first-year PhD student and started in October 2016 at the University of Warwick under the supervision of Professor Chenlei Leng. Previously he studied at the University of Auckland, NZ and University of Malaya, Malaysia. His background is in Mathematics and Statistics and his current research interest includes sports data analysis, networks and machine learning from statistical point of view.

## Research

In a sports league, each team may face each other a finite number of times. Based on the results, we will be able to come out with a model that predicts the winning probability of a team against another specific team, for example by using the Bradley-Terry model. It is of interest to generalise this model to include interaction terms that may affect the outcome of the match, for example whether the team is playing as home or visitor. In his research, Hilmi will be focusing on this type of problem and extending it for more complex cases.

# Nicole Peinelt

## Warwick

## Bio

Nicole Peinelt is currently a first year doctoral student at the Department of Computer Science at the University of Warwick and The Alan Turing Institute with a special interest in Natural Language Processing. Prior to that, she studied Chinese language and Computational Linguistics in Germany, China and Taiwan.

## Research

Nicole’s research interests include information retrieval and computer-assisted language learning. Her doctoral work explores machine learning algorithms for community-based question answering and text summarisation. She is particularly interested in applying these methods for real-world educational purposes.

# Odysseas Sclavounis

## Oxford

## Bio

Odysseas is a graduate of King’s College London and of the University of Oxford. He recently completed his masters degree at the Oxford Internet Institute. His thesis was on reputation stability on dark net marketplaces. Odysseas has now progressed onto the DPhil program at the Oxford Internet Institute under the supervision of Vili Lehdonvirta.

## Research

At The Alan Turing Institute, Odysseas will be studying how public blockchains such as Bitcoin are governed. The ongoing blocksize debate and the DAO fallout, show how contentious the governance of these socio-technical systems has been.

# Shouto Yonekura

## UCL

## Bio

Shouto studies at the Department of Statistical Science, at University College London. His current research interests are statistical methods for stochastic differential equations with discrete observations.

## Research

Shouto’s current research interests include studying model selection methods for diffusion-type models with discrete time observations. Although empirical studies using diffusion-type models have been exponentially increasing, there are quite a few studies on model selection methods for these models. This is basically because the likelihood functions of diffusion-type models with discrete time observations are intractable. To overcome this difficulty, Shouto’s research seeks to develop Sequential Monte Carlo methods (a special case of Feynman-Kac Formulae) to make it possible to construct model selection methods for diffusion-type models with discrete time observations. He is also interested in Sequential Monte Carlo methods for parameter estimation.

# Vidhi Lalchand

## Cambridge

## Bio

Vidhi is a first year PhD student at the University of Cambridge, Department of Physics, Laboratory for Scientific Computing.

Prior to this she completed a MPhil in Scientific Computing from Cambridge (2016) and a M.Sc. in Applicable Mathematics from the London School of Economics and Political Science (2010) securing Distinction in both. Before joining Cambridge in 2015 she worked as a quantitative analyst at Credit Suisse and as a high frequency trader at the Chicago based hedge fund, Citadel Securities (Europe) in London between 2011 and 2015. Her supervisors are Dr. Anita Faul and Dr. Christopher Lester from Cambridge.

## Research

Her research can be broadly categorised as supervised machine learning and its applications. On the theoretical side she is developing a non-parametric framework for supervised learning which uses Bayesian inference. Bayesian inference is a sub-field of machine learning which entails learning from data in a way that is inherently probabilistic. The result of the “learning” exercise is to make predictions on unseen data in a continuous context (regression) or categorical labels (classification). Prediction in Bayesian inference returns a probability distribution rather than a single point estimate, in this sense the former accounts for uncertainty in predictions.

A measure of statistical guarantee or confidence in machine learning predictions is becoming an increasingly important requirement for their usage in several industrial applications where the cost of false positives is really high. For instance, in medicine where the task is to decide whether to operate on a patient or not based on a classification exercise of cancerous concentrations of cells, the probability of a correct prediction is extremely important. Bayesian nonparametrics is the only current paradigm that allows the user to stipulate a posterior belief or a prediction in terms of a probability.

She is equally interested in the potential of existing machine learning methodologies to data driven problems in particle physics, astronomy, quantitative finance and medicine. While still in early stages and subject to modifications, she has titled her thesis – “Probabilistic Kernel Machines: Theory and Applications” to encompass the full extent of work under her PhD.

# Daniel Duma

## Edinburgh

## Bio

Daniel is a PhD student at the School of Informatics, University of Edinburgh, now in his 3rd year. His interests are clustered around Natural Language Processing: he has previously worked on Natural Language Generation and Linked Open Data, and he is currently working on NLP applied to Information Retrieval.

## Research

Daniel’s research falls under the title “context-based citation recommendation”: automatically recommending citations to authors of academic papers. His approach attempts to connect statements in a draft paper with similar statements found in a collection of previous literature, with the aim to find support for these and more importantly help identify relevant research. He has so far incorporated argumentation schemes such as Core Scientific Concepts and Argumentative Zoning to this task and is currently working on the automatic extraction of keywords and generation of queries from the draft paper using linguistic features.

# Chanuki Seresinhe

## Warwick

## Bio

Chanuki Illushka Seresinhe is a third year doctoral researcher from the Data Science Lab, Warwick Business School. Chanuki’s research interest entails using online data from such sources as Flickr and Google Street View to understand how the aesthetics of the environment impacts human wellbeing. Prior to returning to university, Chanuki worked in the digital design sector for over ten years.

## Research

Using data from Scenic-Or-Not, a website that crowdsources ratings of “scenicness” of 1 km grid squares of Great Britain, combined with data from the Census, we find that residents of more scenic environments report better health (see Quantifying the Impact of Scenic Environments on Health, Seresinhe, Preis & Moat, 2015) even when taking core socioeconomic indicators, such as income, into account.

However, visual environments can vary dramatically throughout 1km grid squares, especially in cities. This research project will apply deep learning techniques to images of urban locations, in order to explore which visual features of the environment might impact the well being of city residents.

# Bingyu Zhao

## Cambridge

## Bio

Bingyu Zhao is a third year PhD student from Cambridge University. From October 2016 to September 2017, she is an exchange student at The Alan Turing Institute under the `Enrichment` programme. She is a Civil Engineer and a learner of data science for solving Engineering problems.

## Research

Bingyu’s PhD project studies the degradation of transport infrastructures, focusing on the occurrence and cure of potholes, cracks and other types of defects. The topic itself concerns every road user and is a long-standing problem in Transportation Engineering. Building on previous work, she uses an Agent-based, Bottom-up, City-scale model for degradation simulation and maintenance planning. According to the Data Science Venn Diagram (Drew Conway, 2010), her work combines:

**Domain Knowledge**of the degradation problem and existing studies**Math & Statistics**to build meaningful models**Hacking skills**for analysing data and communicating results

# Pankaj Pansari

## Oxford

## Bio

Pankaj is a second year PhD student in the Engineering Science department at University of Oxford and an enrichment year student at The Alan Turing Institute. His research interests lie in optimisation and machine learning. At Oxford, he is a member of the OVAL group under Prof. Pawan Mudigonda.

## Research

Pankaj’s current project is on the optimisation of convolutional neural network (CNN) architecture. The multi-layer topology of CNNs with hierarchical feature representation has been an important factor for their success in computer vision applications. However, there is still a lack of a principled approach to selecting the right architecture for a given problem. In this project, his research group attempts to present a new optimisation framework for simultaneously learning both the CNN structure and the network weights for a given problem. They aim to later extend the method to recurrent neural networks.

# Nina Otter

## Oxford

## Bio

Nina is a second year DPhil student in mathematics at the University of Oxford, and her supervisors are Ulrike Tillmann and Heather Harrington. She is interested in using methods from pure mathematics (such as algebraic topology, algebraic geometry and category theory) to study data.

## Research

The aim of Nina’s project is to tackle the problem of finding computable invariants for multi-parameter persistent homology. Persistent homology is a method in topological data analysis which allows the study of qualitative features of data across different values of a parameter, which one can think of as scales of resolution, and provides a summary of how long individual features persist across the different scales of resolution. In many applications, data depends not only on one, but several parameters, and to apply persistent homology to such data, one needs to study the evolution of qualitative features across several parameters. While the theory of 1-parameter PH is well-understood, the theory of multi-parameter persistent homology presents many challenges.

# Merve Alanyali

## Warwick

## Bio

Merve Alanyali is a PhD researcher in Data Science Lab at Warwick Business School, University of Warwick with the Chancellor’s International Scholarship. Her research focuses on the analysis of large open data sources with concepts and methods stemming from image processing and machine learning. She graduated with high honours in Computer Science from Izmir University of Economics in 2011, and hold a double-degree MSc. in Complex Systems Science from University of Warwick, UK and Chalmers University of Technology, Sweden.

## Research

Increasing quantities of global data on human behaviour are being generated through our everyday interactions with the Internet. Data gathered from people’s interactions with the Internet tend to be available at speed and low cost, capturing everyday behaviour at a nation or even global scale.

In Merve’s PhD, she is exploiting the potential to study social processes as they unfold by analysing photographic data uploaded to the Internet in near to real time. She draws on her knowledge of image analysis and machine learning methods developed in computer science and statistics to contribute to the computational social science literature by creating new measurements of human behaviour to be of value to social scientists and policy makers alike.

# Marina Riabiz

## Cambridge

## Bio

Marina’s a third year PhD student at the University of Cambridge, Department of Engineering, Signal Processing and Communications group. Her research interest lies in the statistical modelling of heavy-tailed distributions, that describe phenomena with extreme values. The main challenge is the lack of closed-form mathematical expressions and she is following the fascinating Bayesian approach to statistical inference.

## Research

So far Marina has studied the α-stable class of heavy-tailed distributions because of the straightforwardness of its parameters. She has first proposed a method to estimate the latter, based on algorithms for intractable likelihood problems. She has then analysed time-series driven by stable noise, aiming at accurately predicting their jumps. In detail, she is extending existing techniques to the multidimensional case, to include the derivatives in the state space. At The Alan Turing Institute, Marina will study how to do inference for the more generic class of Lévy processes, that include the α-stable one, using Bayesian Non-Parametric methods.

# Luca Melis

## UCL

## Bio

Luca is a third year PhD student at the Information Security Group of the Computer Science department at University College London. He works under the supervision of Dr. Emiliano De Cristofaro, and collaborates with Dr. George Danezis. His research interests are in the fields of applied cryptography, privacy and cloud security. Before joining UCL, he received his BSc and MSc in Computer Engineering from the University of Florence.

## Research

Luca’s current research revolves around privacy-friendly computation aiming to enable computational scenarios and applications where sensitive data could be used to extract useful knowledge, but would otherwise be impossible without clear privacy guarantees. In particular, his research focuses on applications including, but not limited to:

- Training machine learning models based on data gathered from many sources in a privacy-preserving way.
- Collaborative Threat Mitigation via optimal information sharing and clustering techniques.

In this context, crucial challenges such as threat model, efficiency and deployment need to be taken into account. Tackling such challenges constitutes the main research.

# Joe Shaw

## Oxford

## Bio

Joe is a third year PhD candidate and research assistant at the University of Oxford. He studied at the University of Manchester and spent five years working in architectural practice and construction before pursuing a MSc in urban geography at the London School of Economics. Joe’s research interests are focused at the intersection of urban geography, real estate and critical data studies.

## Research

Joe’s DPhil research is examining disruption and innovation within the PropTech industry. In particular, he is interested in the market effects of platforms that promise heightened accuracy and efficiency in the property valuation process. His research seeks to explain how such technologies are transforming property market practices, and how this is affecting different processes and spaces of urbanisation.

# Jeremy Reizenstein

## Warwick

## Bio

Jeremy is in his third year of his PhD at Warwick’s Centre for Complexity Science. Before that, he worked in mathematical finance. Interests: handwriting recognition, recurrent neural networks, rough paths, financial derivatives, software engineering.

## Research

Jeremy’s topic is in improving deep learning of data which contains a time dimension. He looks at online Chinese handwriting recognition, where he wants to write a machine-learning algorithm (a computer program) which should ‘learn’ by example to identify a Chinese character which someone has drawn on a touchpad. Iterated-integral signatures are a recent mathematical encoding of the general shape of a path. They could be used as a step in the operation of a machine learning procedure to reduce the computational effort required to get good accuracy, and this is what he is working with.

# James Pitkin

## UCL

## Bio

James is a 3rd year PhD student based at UCL’s Statistics department. His research looks at the application of Bayesian nonparametric methodologies to retail analytics. Prior to the PhD he worked in the City and a tech start-up. He holds a BA in Mathematics from Cambridge University. He’s interested in the areas of forecasting, clustering, out-sample-prediction and mixture modelling.

## Research

James’ project aims to develop demand models for sparse count processes in retail analytics. Products with low sales volumes are known as “slow-moving goods” in the retail analytics field and are recognised to be difficult to model using standard regression tools. His project looks to combine zero-inflated modelling with self-exciting features within a Bayesian non-parametric framework, such that groups of products can be described by the same process. This allows information borrowing as well as sales characterisation across different products. He implements the methods on a dataset provided by industrial partner, Dunnhumby Ltd.

# Hasiba Afzalzada

## UCL

## Bio

Hasiba is a third year PhD student at University College London. She is interested in security modeling, specifically in modeling the behaviour of security systems in the presence of different types of attackers. She is also interested in the investment strategies made to aid the process of decision making to *effectively* improve security.

## Research

Hasiba is currently working on the analysis of the behaviour of a security system in the presence of attackers. The system’s architecture model is represented by the methodology inspired by the classical distributed systems and the behaviour of attackers is modelled using Markov Decision Processes. These models can be embedded in Impulse Response Functions, which take into account the three security attributes Confidentiality, Integrity and Availability, to reliably show how a system restores itself after facing an attack. Findings will be applied to real life examples, such as military or retail, to obtain an effective way of investing in security.

# Elena Kochkina

## Warwick

## Bio

Elena is a second-year PhD student in Computer Science at the University of Warwick supervised by Professor Rob Procter and Dr. Maria Liakata. Her research interests are Machine Learning, Data Mining, Deep Learning, Natural Language Processing, Social Media, Multi modal Learning.

## Research

Elena’s PhD project is focused on integrating text and image data for content verification and opinion mining in social media. The motivation for this is the fact that in recent years images in social media (e.g. Twitter) have become a very powerful tool to transfer information and express emotions. Her research seeks to combine information extracted from images and text to gain insights on people’s stance towards various issues and to understand phenomena such as rumour spread and opinion formation. She is interested in the application of Recurrent Neural Networks and Multi modal Distributional Semantic Models for this purpose.

# Sebastian Bobadilla-Suarez

## UCL

## Bio

Sebastian is a fourth year PhD student at University College London studying Experimental Psychology. His research interests include bridging machine learning with cognitive neuroscience, both with respect to methods and with respect to theory. He has studied heuristics, control, decision-making, categorization and neural similarity. His background is in Social Anthropology with fieldwork in various locations in Mexico.

## Research

The project aims to understand how the brain codes information states by determining what makes two brain states similar. Using various data sets, various similarity measures will be compared to a benchmark to determine the best measure. To derive the benchmark, various classifiers will be applied to the neural data with the class labels corresponding to what the person was viewing during the brain scan. The best classifier’s confusion matrix will serve as the benchmark to evaluate the similarity measures.

# Rodrigo Mendoza-Smith

## Oxford

## Bio

Rodrigo Mendoza-Smith is fourth year D.Phil. student at the Mathematical Institute in the University of Oxford. He received a BSc in Applied Mathematics from the Mexican Autonomous Institute of Technology (ITAM) in 2012 and the MSc in Mathematical Modelling and Scientific Computing from the University of Oxford in 2013.

## Research

Rodrigo’s research interests lie in the intersection of Information Theory and Geometry, and have a focus on Numerical Computing and Machine Learning. At The Alan Turing Institute, Rodrigo will focus on developing algorithms for Persistent Homology computation and Topological Data Analysis.

# Valerio Giuffrida

## Edinburgh

## Bio

Valerio is a PhD student in Computer Science. His background encompasses mainly machine learning and pattern recognition applied to images. He is in his last year of the program at the University of Edinburgh. As a researcher in computer vision, he has focused his studies on neural networks learning suitable image representation for the task at hand.

## Research

The goal of his project is to devise novel approaches to learning compact invariant representations. Specifically, he will extend his recently published work on Rotation-Invariant RBM in the context of deep learning, in order to learn higher-level features that have been proven to be successful for several applications, compared to shallow networks. The first step is to use ERI-RBM within a Deep Belief Network. Another direction he considers important is to be able to demonstrate benefits in real applications beyond ones considered: counting multiple instances of the same object in an image.

# Yiwei Zhou

## Warwick

## Bio

Yiwei is a third-year PhD student from Department of Computer Science at the University of Warwick, supervised by Dr. Alexandra I. Cristea. His research interests are Data Mining, Machine Learning and Natural Language Processing.

## Research

Yiwei’s PhD project is about mining people’s views on entities/events in social media. Former projects include analysing sentiment and topic differences in multilingual Wikipedia on entities/events, summarising real-time timeline for high-impact events in Twitter, etc.

# Thibaut Lienart

## Oxford

## Bio

Thibaut is a 4th year PhD student in computational statistics co-supervised by Professor Arnaud Doucet and Yee Whye Teh. Thibaut is affiliated with University College in Oxford. Prior to starting the PhD, Thibaut took part iii in Cambridge and before that obtained a Masters in applied mathematics from UCLouvain in Belgium. Research interests include variational inference and particle methods applied on graphical models and in the context of distributed learning.

## Research

Thibaut’s research covers inference on graphical models with applications to large-scale machine learning problem. Graphical models permit to represent conditional dependence structure between a network of random variables and can be used to represent a wide range of machine-learning techniques. However performing inference on them can be hard especially if the potentials are continuous and non-gaussians. Thibaut has so far looked at exploiting particle methods techniques as well as variational inference techniques and the combination of the two to deal with inference on graphical models. Recently he has also been looking at rejection-free sampling methods and their potential application on graphical models.

# Abhinav Mehrotra

## UCL

## Bio

Abhinav is a third year computer science PhD student at the University of Birmingham. He works under the supervision of Mirco Musolesi and Robert Hendley. His main areas of interest include social computing, context-aware computing, digital health and human behaviour modelling.

## Research

Abhinav’s current research focuses on understanding and predicting human behaviour for interacting with mobile devices by using contextual information obtained from the embedded sensors. He also work on the projects related to monitoring and predicting the cognitive context (such as mood) as well as the health of people through the analysis of mobile phone data.

# Adam Tsakalidis

## Warwick

## Bio

Adam is a third-year PhD candidate in Urban Science with the Warwick Institute for the Science of Cities, supervised by Dr. Alexandra I. Cristea and Dr. Maria Liakata. His main research interests are in the areas of Social Media Mining and Natural Language Processing, with a primary focus on sentiment analysis from social media and their implications in real-world events.

## Research

Over 70% of the global population is expected to live in big cities by the next 30 years. At the same time, the use of social media is growing at tremendous rates. Motivated by these basic facts, Adam’s work focuses on opinion mining of social media users, as an attempt to model and predict urban-related indices (e.g., elections, wellbeing, etc.). The effective modelling of such domains is a challenging and important task, as it can provide additional “sensors” for informing citizens and policy makers on the current state of an urban environment, but also on possible consequences of their actions.

# Alkeos Tsokos

## UCL

## Bio

Alkeos is currently in his second year of a PhD at the Statistics Department of UCL, supervised by Dr Ioannis Kosmidis. Prior to the PhD he completed an MSc in Statistics, followed by an MSc in Machine Learning, both at University College London.

## Research

Alkeos’ undergraduate project, supervised by Dr Yvo Pokern, examined goodness of fit methods for stochastic differential equations. For his MSc, his dissertation was completed at the offices of MoneySuperMarket.com where he looked at regularisation methods for multinomial logistic regression models applied to conversion rate modelling. Alkeos’ PhD project is part funded by the English Institute of Sport and broadly the purpose is to apply statistical techniques to answer questions sports practitioners and coaches may be interested in. In particular, they are exploring the potential of functional data analysis to analyse data arising from wearable sensors worn by athletes during training

# Aristeidis Panos

## UCL

## Bio

Aristeidis is a second year PhD student at the department of statistical science at University College London (UCL). After a four-year programme on informatics at Athens University of Economics and Business, he joined UCL to pursue a PhD on statistics. His research interests lie mostly on the computation of Gaussian processes and their applications in various fields using Bayesian statistics.

## Research

Aristeidis’ project is mainly focused on the efficient computation of Gaussian Processes. Gaussian Process models are widely used to perform Bayesian non-linear regression and classification – tasks that are central to many machine learning problems. One of the fundamental issues of Gaussian Processes is the computational difficulties of applying them to large data sets. Hence, the main goal of the project is to find a way to reduce the computational complexity of Gaussian Processes by using the property of Givens angles to parameterize any eigenvector matrix. This means that essentially, Aristeidis needs to prove that there is a functional relationship between the Givens angles and the elements of the Gaussian Process’ covariance matrix.

# Tejas Kulkarni

## Warwick

## Bio

Tejas is a 2nd year PhD student with Professor Graham Cormode at Warwick University. He holds a Masters in Computer Science from the Indian Institute of Technology, Madras, India. He gained experience as a software engineer in India, before joining the EPSRC CDT in Urban Science in 2015.

## Research

His research interests are based on the theory and applied area of differential privacy which, broadly speaking, aims to maximise query accuracy from data whilst preserving anonymity. Such techniques are crucial for the trusted statistical analysis of sensitive city-scale information, such as voter registration information, medical records or email analysis.

Webpage – Tejas Kulkarni