Speakers

Academic Keynote Speaker

René Vidal, Johns Hopkins Mathematical Institute for Data Science

Title and Abstract:

Semantic Information Pursuit

In 1948, Shannon published a famous paper, which laid the foundations of information theory and led to a revolution in communication technologies. Critical to Shannon’s ideas was the notion that a signal can be represented in terms of “bits” and that the information content of the signal can be measured by the minimum expected number of bits. However, while such a notion of information is well suited for tasks such as signal compression and reconstruction, it is not directly applicable to audio-visual scene interpretation tasks because bits do not depend on the “semantic content” of the signal, such as words in a document, or objects in an image. In this talk, I will present a new measure of semantic information content called “semantic entropy”, which is defined as the minimum expected number of semantic queries about the data whose answers are sufficient for solving a given task (e.g., classification). I will also present an information-theoretic framework called ``information pursuit'' for deciding which queries to ask and in which order, which requires a probabilistic generative model relating data and questions to the task. Experiments on handwritten digit classification show, for example, that the translated MNIST dataset is harder to classify than the MNIST dataset. Joint work with Aditya Chattopadhyay, Benjamin Haeffele and Donald Geman.

 

Defence Keynote Speaker

Hugh Griffiths – Defence Science Expert Committee (DSEC) / University College London

Title and Abstract:

Why Defence Acquisition is Difficult

The acquisition of defence equipment – whether tanks, aircraft carriers or fast jets – is notoriously difficult.  On the occasions when it goes wrong we hear of cancelled projects or costly spending overruns, amounting to billions of pounds of taxpayers’ money. This is not just a UK problem – it happens in other countries too. 
This presentation looks at the process of defence acquisition, and particularly the research and development that underpins it. Examples are given of both successes and failures. In particular, it examines the nature of innovation and of low-TRL research and considers the appropriate balance between risk and ambition.

 

 

Invited Speakers

Mark Briers, The Alan Turing Institute

Title and Abstract:

Optimising and Understanding the Impact of the NHS COVID-19 app Using Data Science

The NHS COVID-19 app has been downloaded over 21 million times. It was the first app in the world to use the new Google / Apple  API, allowing novel Bayesian statistical methods to infer distance. Our latest research demonstrates that the app has reduced the number of cases by approximately 600,000 in 2020. The app data are providing valuable insights into the evolution of the pandemic across England and Wales. This presentation will provide an overview of the NHS COVID-19 app, the impact results, and our potential future directions.

 

Tien Pham,  (CISD) U.S. DEVCOM ARL

Title and Abstract:

AI-enabled Multi-Domain Processing and Analytics for Decision Making

The U.S. Army is evolving its warfighting doctrine to militarily compete, penetrate, disintegrate, and exploit peer adversaries in a Multi-Domain Operations (MDO) Joint Force. To achieve this vision, the Army requires Artificial Intelligence (AI) and Machine Learning (ML) capabilities to enable analytics from the full spectrum of multi-domain data for decision making by highly-dispersed teams of humans and robot agents. This highly diverse learning and reasoning span data types (e.g. video, open-source multi-media, electromagnetic (EM) signals, sensors), warfighting domains (e.g. land, air, cyberspace, spectrum), and warfighting functions (e.g. intelligence, command and control (C2), fires, protection). MDO presents unique challenges that must be overcome by any AI/ML application, including forwarding deployment in complex terrain, dynamic, distributed, resource-constrained environments, and highly contested settings. Moreover, these applications must consider the inherent uncertainties (model and environmental) and the operational timeliness requirements (i.e., time available to learn, infer, and act).

The presentation will discuss the U.S. Combat Capabilities Development Command (DEVCOM) Army Research Laboratory (ARL) collaborative, cross-cutting AI research efforts for MDO to enable processing and analytics for decision making toward the tactical edge. The research efforts will address the technical challenges involving (i) learning and reasoning in complex MDO environments and (ii) resource-constrained AI processing at the point of need.

 

Alan Hunter, University of Bath

Title and Abstract:

Multi-Spectral and Multi-Modal Underwater Acoustic Imaging

The useable spectrum of underwater acoustic frequencies is very broad, spanning roughly 6 or 7 orders of magnitude from Hz up to MHz. Traditionally, acoustic imaging of the seafloor has been carried out at high carrier frequencies (in the range of 100 kHz to 1 MHz) using relatively narrow-band signals (quality factors much greater than 1). This has enabled the production of high-resolution acoustic images that often resemble optical photographs. However, there are opportunities to exploit much more of the spectrum and thereby access a richer source of acoustic information. An effective means of achieving this is to operate in multiple separate spectral bands. Bands in the higher frequencies retain the advantages of fine resolution, differences between bands can be used to measure frequency-dependent scattering characteristics, and lower frequency bands (below 100 kHz) increasingly enable penetration beneath the sediment and inside objects.  Moreover, low-frequency acoustic waves can couple with elastic wave modes in solids and this gives the potential for sensing material composition and structure for enhanced characterisation.  This talk will explore the opportunities and challenges of multi-spectral and multi-modal (i.e., acoustic and elastic) imaging with regard to signal processing, image visualisation, and interpretation.