PhD studentship (UK home students), University College London Deadline: 28th of April, 2023, closing of business hours (5pm BST)
A fully funded PhD studentship covering home fees and a stipend of about £22K for 4 years is available from a starting date in September 2023 (please notice that this does not cover students who do not qualify as UK home students). The student will be supervised by Prof Ricardo Silva, from the Department of Statistical Science UCL and partially funded by dunnhumby. A description of the project is provided below. The successful candidate will join the Centre for Doctoral Training (CDT) in Foundational Artificial Intelligence hosted by the Computer Science department (https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ucl.ac.uk%2Ffoundational-ai-cdt%2Ffoundational-artificial-intelligence-mphilphd&data=05%7C01%7Cuai%40engr.orst.edu%7Ca1586e6b7d7c4ca5857d08db25676db7%7Cce6d05e13c5e4d6287a84c4a2713c113%7C0%7C0%7C638144898376836492%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=kH5YsZQR2Zl8o5HfoD2ekdCBuWl0W2UGruv41qgpBX8%3D&reserved=0). A unique aspect of the CDT is to give students the deep technical skills they require to be leading researchers in AI and also the skills to be a deep tech entrepreneur. As a member of the CDT, you will be expected to actively engage on cohort-based training activities designed to encourage interaction with other students on the CDT, through shared research, training and social experiences, including in AI and entrepreneurship. How to apply: via https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ucl.ac.uk%2Ffoundational-ai-cdt%2Fhow-apply&data=05%7C01%7Cuai%40engr.orst.edu%7Ca1586e6b7d7c4ca5857d08db25676db7%7Cce6d05e13c5e4d6287a84c4a2713c113%7C0%7C0%7C638144898376836492%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=YNbgoAezmzlM3yYRk47%2BcpKPvFuzLO6AZbamiHMOujo%3D&reserved=0 , mentioning in your application documents that you are interested in this particular project. Please notice that the deadline for this specific studentship is April 28th 2023, 5pm (BST). Topic: Causality in retail data science It is an often-repeated phrase in artificial intelligence (AI) and data science that ‘association does not imply causation’. It is the ability to perform causal reasoning that is often recognised as one of the hallmarks of human intelligence. However, traditional statistical, AI, machine learning, and data science methods are based on association. In recent years, developments in the field of causality have provided scientists with formal tools and a mathematical apparatus for moving beyond association, to causation. One key approach for establishing causation is randomised controlled trials, which however are not always possible or feasible to perform in practice, in which case observational data is used instead. For observational data, the most widely adopted framework of causal inference in computer science is given by the structural causal models, graphical causal models and do-calculus, championed by Professor Judea Pearl (Turing Award, 2011). The work published by Prof Pearl and his collaborators on this topic is seen by many as heralding the start of a ‘causal revolution’. This causal framework allows researchers to specify causal assumptions and derive theoretical guarantees, which allow for the use of observational data to answer “what if” questions about interventional or counterfactual (unobserved) outcomes. Further benefits include the intrinsic explainability and transparency of these causal models, which are important features given the repeated calls for transparent and explainable AI, and the rapid development of this field, in recent years. Whilst the interest of commercial companies in causality has grown, the exploration of practical applications in the retail domain is only now beginning. Given the great wealth of data available in retail about real-world customer transactions happening online and offline over time, investigating how causal methods can be applied to retail is very timely and promising. As a leader in retail data science, dunnhumby has access to the most comprehensive and granular data available together with an unparalleled understanding of the retail sector. This project will investigate the real-world application of causal methods to typical retail use cases, using data commonly available to retailers. There is a variety of causal use cases at dunnhumby, ranging from observational to randomised experimental studies, using online or offline data, estimating different types of causal effects depending on the context each time. These use cases may include examples such as: measuring the impact of media campaigns; predicting the effects of range changes; understanding what drives customer loyalty. The goals of the project include: demonstrating that it is possible to apply causal methods to complex real-world problems given data and experimental set-up constraints; demonstrating the benefits of such applications of causal methods in retail; and ultimately creating a blueprint for their future use.
_______________________________________________ uai mailing list uai@engr.orst.edu https://it.engineering.oregonstate.edu/mailman/listinfo/uai
