Overview: We propose a Centre for Doctoral Training in Data Science. Data science is an emerging discipline that combines machine learning, databases, and other research areas in order to generate new knowledge from complex data. Interest in data science is exploding in industry and the public sector, both in the UK and internationally. Students from the Centre will be well prepared to work on tough problems involving large-scale unstructured and semistructured data, which are increasingly arising across a wide variety of application areas. Skills need: There is a significant industrial need for students who are well trained in data science. Skilled data scientists are in high demand. A report by McKinsey Global Institute cites a shortage of up to 190,000 qualified data scientists in the US; the situation in the UK is likely to be similar. A 2012 report in the Harvard Business Review concludes: "Indeed the shortage of data scientists is becoming a serious constraint in some sectors." A report on the Nature web site cited an astonishing 15,000% increase in job postings for data scientists in a single year, from 2011 to 2012. Many of our industrial partners (see letters of support) have expressed a pressing need to hire in data science. Training approach: We will train students using a rigorous and innovative four-year programme that is designed not only to train students in performing cutting-edge research but also to foster interdisciplinary interactions between students and to build students' practical expertise by interacting with a wide consortium of partners. The first year of the programme combines taught coursework and a sequence of small research projects. Taught coursework will include courses in machine learning, databases, and other research areas. Years 2-4 of the programme will consist primarily of an intensive PhD-level research project. The programme will provide students with breadth throughout the interdisciplinary scope of data science, depth in a specialist area, training in leadership and communication skills, and appreciation for practical issues in applied data science. All students will receive individual supervision from at least two members of Centre staff. The training programme will be especially characterized by opportunities for combining theory and practice, and for student-led and peer-to-peer learning.

Traditionally, most software projects have been tackled using a single programming language. However, as our ambitions for software grow, this is increasingly unnatural: no single language, no matter how "good", is well-suited to everything. Increasingly, different communities have created or adopted non-traditional languages - often, though not always, under the banner of Domain Specific Languages (DSLs) - to satisfy their specific needs. Consider a large organisation. Its back-end software may utilise SQL and Java; its desktop software C#; its website back-end PHP and the front-end Javascript and HTML5; reports may be created using R; and some divisions may prototype software with Python or Haskell. Though the organisation makes use of different languages, each must execute in its own silo. We currently have few techniques to allow a single running program to be written using multiple languages. In the Cooler project, we call this the "runtime composition" problem: how can languages execute directly alongside each other, exchange data, call each other, optimise with respect to each other, etc.? The chief existing technique for composing language runtimes is to translate all languages in the composition down to a base language, most commonly the byte code for one of the "big" Virtual Machines (VMs) - Java's HotSpot or .NET's CLR. Though this works well in some cases, it has two major problems. Firstly, a VM will intentionally target a specific family of languages, and may not provide the primitives needed by languages outside that family. HotSpot, for example, does not support tail recursion or continuations, excluding many advanced languages. Secondly, the primitives that a VM exposes may not allow efficient execution of programs. For example, dynamically typed languages running on HotSpot run slower than their seemingly much less sophisticated "home brew" VMs. The Cooler project takes a new approach to the composition problem. It hypothesizes that meta-tracing will allow the efficient composition of arbitrary language runtimes. Meta-tracing is a recently developed technique that creates efficient VMs with custom Just-in-Time (JIT) compilers. Firstly, language designers write an interpreter for their chosen language. When that interpreter executes a user's program, hot paths in the code are recorded ("traced"), optimised, and converted into machine code; subsequent calls then use that fast machine code rather than the slow interpreter. Meta-tracing is distinct from partial evaluation: it records actual actions executed by the interpreter on a specific user program. Meta-tracing is an exciting new technique for three reasons. Firstly, it leads to fast VMs: the PyPy VM (a fully compatible reimplementation of Python) is over 5 times faster than CPython (the C-based Python VM) and Jython (Python on the JVM). Secondly, it requires few resources: a meta-tracing implementation of the Converge language was completed in less than 3 person months, and runs faster than CPython and Jython. Third, because the user writes the interpreter themselves, there is no bias to any particular family of languages. The Cooler project will initially design the first language specifically designed for meta-tracing (rather than, as existing systems, reusing an unsuitable existing language). This will enable the exploration of various aspects of language runtime composition. First, cross-runtime sharing: how can different paradigms (e.g. imperative and functional) exchange data and behaviour? Second, optimisation: how can programs written in multiple paradigms be optimised (space and time)? Finally, the limits of the approach will be explored through known hard problems: cross-runtime garbage collection; concurrency; and to what extent runtimes not designed for composition can be composed. Ultimately, the project will allow users to compose together runtimes and programs in ways that are currently unfeasible.

Computer systems have been pervasive for many years, but despite this, and despite the huge resources devoted to their construction, they are still typically insecure, prone to failure, and hard to use. Major failures are commonplace, in sharp contrast with the products of other engineering industries, and dealing with them, and with the day-to-day lesser flaws, has huge economic and social costs. The core technical difficulty is system complexity: the range of behavior, the large scale, and the legacy of old design choices combine to make it hard to understand these systems well enough to engineer them well. My main research goal is to develop intellectual tools that suffice for solid system-building, analogous to the applied mathematics of more traditional engineering disciplines. This must be grounded on real systems - it cannot be done in theoretical isolation. My approach, as documented in the Track Record, is to focus on the key articulation points in the hierarchy of abstractions used to build systems: programming languages, processor instruction sets, network protocols, and so forth. These are relatively stable points in a rapidly changing environment, are critical to all system development, and are small enough that a modest team can address them. Each demands different research: new language constructs, new specification, reasoning, and testing techniques, and so forth. In this Fellowship I will pursue this approach, focussing on the problems in building computer systems above the intricate relaxed memory models of modern multiprocessors. Multiprocessor systems are now the norm (as further speed-up of sequential processors has recently become impractical), but programming them is very challenging. A key difficulty is that these systems do not provide a sequentially consistent memory, in which events appear to occur in a single global time order, but instead permit subtle reorderings, rendering intuitive global-time reasoning unsound. Much previous work across a range of Computer Science, in programming languages, program logics, concurrency theory, model checking, and so on, makes the now-unrealistic assumption of sequential consistency, and it must now be revisited in this more complex setting.I will develop precise mathematical models of the behavior of real-world multiprocessors that take such reorderings into account, and develop semantics and reasoning techniques above them. Using those, I will consider the verification of high-performance concurrent algorithms (as used in operating system and hypervisor kernels), the design of higher-level languages, and verified compilation of those languages to real machines. This will enable future applications to be developed above a high-confidence and high-performance substrate. It should also have a broader beneficial effect on research in Computer Science, drawing together mathematically well-founded theory and systems-building practice.

In 2003. over 44,000 people in the UK were diagnosed with breast cancer. This is now the commonest cancer occurring in the UK. The lifetime risk of developing breast cancer is 1 in 9, and while most of the women who get breast cancer are past their menopause, almost 8,000 diagnosed each year are under 50 years old. Improving outcomes is a key challenge in treatment. High throughput genomic methods, such as expression profiling of frozen tissue samples using microarrays, have resulted in the discovery of many novel gene signatures that are correlated with clinical outcomes in cancer treatment. It is essential that these potential biomarkers are validated on large numbers of independent samples prior to clinical use. Tissue microarrays (TMA) created from paraffin-embedded tumour samples from large clinical trials are the ideal reagent for these validation experiments. However, the analysis and scoring of antibody-based markers on TMAs that may contain thousands of patient samples presents major challenges for pathology reporting and image handling. Our initial pilot PathGrid study (Oct 2007 to Oct 2008) is using a range of techniques which have been developed in astronomy to both analyse imaging data and to handle and manipulate the resulting data products. These have been applied to the challenges involved in analysing the TMA image data taken from the SEARCH study population-based study of breast cancer. We have utilised astronomy 'Virtual Observatory' components specifically adapted from those developed within the AstroGrid Virtual Observatory eScience programme (http://www.astrogrid.org), to facilitate secure data transport, resource discovery through appropriate metadata, data acquisition, ingression to a database system, and secure distributed access to those data and information resources. Image analysis has been applied to the input TMA data utilising a range of algorithms originally developed for diverse astronomical use cases. The resulting data products will in turn be interfaced (though work planned here) to the clinical trials systems developed through the CancerGrid system (see http://www.cancergrid.org). In our initial test case we have automated the scoring of Estrogen Receptor (ER) assessments. ER is an important regulator of mammary growth, but is also a key prognostic and therapeutic target in breast cancer. Assessing ER status at time of diagnosis of breast cancer, determines which treatment programmes should be followed by patients. In particular, those patients who have ER-positive breast cancer will be offered estrogen antagonist therapies such as tamoxifen. At CR-UK, breast cancer studies utilising genomics tools are underway to validate existing and new prognostic and/or predictive markers. TMAs have been created from a large population-based clinical trial (SEARCH; part of the Anglia Breast Cancer study) for analysis with a range of candidate markers. Immunohistochemistry is used to assess the level of nuclear ER expression. The focus of our initial pilot has been on the algorithm development and validation. For the miniPIPSS programme the work will move to increasing the utility of the analysis system by running all processing operations in a pipeline. This automation, the operational basis of our Pathgrid system, has been implemented in prototype by making use of the application-grid infrastructural components from AstroGrid This miniPIPSS project will facilitate the further interchange of ideas and technologies between the physical and medical sciences, and provide methods for the handling and processing of clinical image data in an open and extensible manner The interaction with Oracle represents the initial stage of a longer term partnership, aiming to further develop these analysis techniques such that they are available to he wider medical research community - and further ahead for use in a clinical environment.

Most modern computer applications depend in some way or another on computations that are performed by server applications on the internet. More and more of these server applications are now built as so-called microservices, which allow developers to gradually update or fix issues in unrelated parts of a larger application, and therefore, have become popular. Many of these microservices avoid certain types of concurrency issues by design. Unfortunately, they still suffer from other kinds of concurrency issues, for example when multiple online customers try to reserve the same seats at the same time. For software engineers, it is hard to test for all possible concurrent interactions. In practice, this means that only simple concurrency issues are reliably detected during testing. Complex issues can however easily slip through and make it into server applications and then handle client requests incorrectly. One example of such a concurrency issue appeared at Nasdaq when the Facebook stock was traded for the first time, resulting in the loss of millions of dollars. Our goal is to develop techniques that detect concurrency issues automatically at run time, to be able to circumvent them, and enable developers to fix them, using detailed information gathered by the detection techniques. Researchers have shown that one can detect and avoid issues, for instance by changing the order in which client requests are processed. In practice however, current techniques slow server applications down significantly, which make these techniques too costly to be used. Our aim is to dynamically balance the need for accurate information and minimize slow down. We conjecture that we can get most practical benefits while only rarely tracking precise details of how program code executes. In addition to automatically preventing concurrency issues to cause problems, we will also use the obtained information to provide feedback to developers so that they can fix the underlying issue in their software. Thus, overall the goal of this research project is to make server applications, and specifically microservices, more robust and resilient to software bugs that are hard to test for and therefore typically remain undiscovered until they cause major issues for customers or companies. Our work will result in the development of adaptive techniques that detect concurrency issues, and automatically tradeoff accuracy and run-time overhead, to be usable in practice. Furthermore, the detection techniques will be used to provide actionable input to the software developers, so that the concurrency issue can be fixed and therefore be prevented reliably in the future. To evaluate this work, we will collect various different types of concurrency issues and make them openly available. This collection will be based on issues from industrial systems and derived from theoretical scenarios for highly complex bugs. We include these theoretical scenarios, since such complex bugs are hard to diagnose and test for, they likely remain undiagnosed and undocumented in practice, but have the potential of causing major disruptions. Finally, we will build and evaluate our proposed techniques based on a system designed for concurrency research. The system uses the GraalVM technology of Oracle Labs, which allows us to prototype at the level of state-of-the-art systems, while keeping the development effort manageable for a small team.