The workshop 'Data analysis and stochastic control: where do statistics and applied probability come together?' was held online on the afternoon of Wednesday 9 December 2020. Organised by the Applied Probability Section, the meeting was attended by an international audience of around 70 academics and practitioners with background mainly in statistics and probability. There were a good number of junior academics and PhD students and the workshop lasted about three hours.
A total of four talks (collectively lasting 1.5 hours) were delivered by Sam Cohen (Oxford), Beatrice Acciaio (ETH Zurich), Blanka Horvath (King's College) and John Moriarty (Queen Mary University), followed by a lively panel discussion with questions from the audience being addressed by the four speakers.
The talks covered mathematical and computational aspects of the use of data in stochastic control. Topics covered included theoretical questions concerning filtering techniques and generative adversarial networks; applied aspects of deep learning models for financial markets; the development of an AI competition environment for the application of machine learning techniques; and the operation of critical infrastructure in the UK (eg, electricity, gas, water and transportation).
The panel discussion revolved around various issues, including if/how theory and practice can be integrated in a coherent modelling framework for stochastic control and machine learning; overfitting of data; stakeholders' understanding of the modelling; and outputs and the role of synthetic data in connection with replicability of the results.