Advances and Challenges in Data Science

Date: 6 June 2017

Time: 10:30-16:30

Venue: Mary Ward House Conference and Exhibition Centre5-7 Tavistock Place, London WC1H 9SN

Click here to book your space. Registration deadline: Registration has now closed

Attendance is free and advance registration is essential. Places can be reserved for the whole day or selected for individual talks.

The talks, themed around ‘Advances and Challenges in Data Science’, range from machine intelligence to the mathematical mysteries of deep networks and public policy. The speakers are drawn from members of The Alan Turing Institute’s Scientific Advisory Board, who are taking part in this unique event during their inaugural visit to the Institute in its first year of research.

10:30 – 11:00 Registration and Coffee

11:00 – 11:45  Limitations on the Practical Adoption of Machine Intelligence

Mike Lynch (Invoke Capital and Council for Science and Technology) 

In his talk, Mike will discuss the limitations on the practical adoption of machine intelligence and the attempts to address them.  He will cover both technical aspects but also real-world issues such as exception processing, legal and insurance.

11:45 – 12:30  Mathematical Mysteries of Deep Networks

Stéphane Mallat (École Polytechnique)

Deep neural networks obtain spectacular classification and regression results over a wide range of data including images, audio signals, natural languages, biological or physical measurements. These architectures can thus approximate a wide range of “complex” high-dimensional functions. This lecture aims at discussing what we understand and do not understand about these networks, for unsupervised and supervised learning.

12:30  – 13:15  Data Science and Public Policy

Robert Devereux (Department for Work and Pensions, UK Government)

Robert will illustrate some of the public policy issues where data science could make a powerful difference, and explore some of the ethical issues that might arise.

13:15 – 14:15   Lunch and Posters 

14:15 – 15:00  Peer Effects, Social Multipliers and Cascades of Human Behaviour: Causal Inference at Scale

Sinan Aral (MIT Sloan School of Management)

In this talk, Sinan will survey empirical approaches to causal inference in networks and describe a series of large-scale randomized experiments and causal observational studies of peer influence to explore the behavioural dynamics catalysed by peer effects or social spill overs in human behaviour and opinion formation. Sinan will discuss the public policy implications of peer effects for bias in online ratings, social advertising, human health interdependence and the ability to generate cascades of behaviour through peer to peer influence in networks

15:00 – 15:45 Three principles for Data Science: Predictability, Stability and Computability

Bin Yu (University of California, Berkeley)

In this talk, Bin will discuss the intertwining importance and connections of three principles of data science in the title in data-driven decisions. Making prediction as its central task and embracing computation as its core, machine learning has enabled wide-ranging data-driven successes. Prediction is a useful way to check with reality. Good prediction implicitly assumes stability between past and future. Stability (relative to data and model perturbations) is also a minimum requirement for interpretability and reproducibility of data driven results (cf. Yu, 2013). It is closely related to uncertainty assessment. Obviously, both prediction and stability principles cannot be employed without feasible computational algorithms, hence the importance of computability.

The three principles will be demonstrated in the context of two neuroscience projects and through analytical connections. In particular, the first project adds stability to predictive modeling used for reconstruction of movies from fMRI brain signals for interpretable models.  The second project use predictive transfer learning that combines AlexNet, GoogleNet and VGG with single V4 neuron data for state-of-the-art prediction performance.  The results lend support, to a certain extent, to the resemblance of these CNNs to brain and at the same time provide stable pattern interpretations of neurons in the difficult primate visual cortex V4.

15:45- 16:30  Afternoon Tea