News and Updates
June - July:
- We are presenting two papers at ICML this year: "Bayesian Autoencoders: Analysing and Fixing the Bernoulli likelihood for Out-of-Distribution Detection" by Xiang, and "Structured Weight priors for convolutional neural networks" by Tim.
- Yaniv will present his paper on "Distributed Dynamic Measures of Criticality for Telecommunication Networks" in the upcoming SOHOMA Conference.
- Liming, Stephen and Alexandra presented our recent work on the AgentChat project in the Connected Everything Conference in July.
- MAG is organising a data study on the UK’s resilience to supply chain shocks in the Agrifood Sector. This study group will take place over two sessions. This first session (7 – 8 September) aims to assess the resilience of the UK food networks, while a second session later in the year will take the representations and learning from session one to look the UK food network response to specific shocks. The event is organized by us, Queen Mary, Knowledge Transfer Network, Isaac Newton Institute for Mathematical Sciences and the Alan Turing Institute. We are particularly interested to hear from researchers in data science, statistical modelling and network science. Please contact firstname.lastname@example.org if you are interested in taking part.
- Dr. Boyang Song from Sheffield, introduced us to his work on “ An Industrial Internet of Things (IIoT) & Deep Learning approach for Predictive Maintenance (PdM)". As part of the Pitch In project we are collaborating with Sheffield to understand how machine learning approaches can be used to improve data challenges in manufacturing quality control.
April - May:
- As the Covid-19 pandemic hit the world, MAG, like most other research groups, has moved to remote working. While our research largely goes on without interruption, MAG is undertaking a number of activities to help combat the crisis:
- First of these is the setting up of an AgriFood Resilience forum with UK’s Mathematical Sciences Research Community. The forum, organised by the Innovate UK and Knowledge Transfer Network will discuss ways in which UK’s Mathematical Sciences community can help combat disruptions to the UK’s food supply and logistics. The forum has published its first report and then gave its first webinar on the 28th April, attracting more than 130 participants. As part of the webinar series MAG presented how Network Science and Modelling can help understand disruption risks in large scale supply systems. Find more details here.
- MAG members have participated in IfM’s Covid-19 taskforce, which includes the development of simulation models to predict patient flows and capacity at Addenbrookes Hospital. Find more about IfM’s initiatives on Manufacturing and Covid-19 here.
- MAG is starting a very timely new project with Aviva on supply chain risk. The project will investigate how rare, catastrophic events can be modelled and how insurance providers can help businesses combat interruption.
- Stephen shared a best presentation award with two other researchers at the annual IfM PhD Conference. Many congratulations to Stephen!
- Edward was selected to participate in the prestigious Max Planck Pre-doctoral School 2020 taking place in Cornell, Maryland.
- Our collaborative research with Cranfield University on Aerospace Ecosystem Networks has been published in PLOS ONE: Jose LA Jr, Brintrup A, Salonitis K (2020) Analysing the evolution of aerospace ecosystem development. PLOS ONE 15(4): e0231985. The paper investigates how network patterns emerge and change over time, comparing countries with developed and emerging Aerospace industrial sectors.
- MAG researchers Liming Xu and Stephen Mak won a feasibility study grant from the UK Connected Everything Network. Our new project, AgentChat, will explore the feasibility of enabling freight co-loading using agent-based technology. Although the idea of freight co-loading has been around since sometime, its implementation has been challenging due to information asymmetry and difficulties involved in manual orchestration: co-located or close suppliers do not know whether they are sending items to similar locations around the same times unless someone tells them. Third-party mediation only works if sufficient agents sign up, and retailers or logistic providers incur transaction costs for orchestration. Our approach will explore how distributed, automated solutions can tackle this problem and what information infrastructures need to be in place for it to be adopted.
- Many congratulations to Edward and Stephen who won first prize in the Aviva Data Science Hackathon organised by the Cambridge University Data Science Society!
- Xiang and Yasmin will be presenting our paper in the IEEE Metrology for Industry 4.0 and IoT conference taking place in Rome in June 2020. The paper details the development of a Bayesian Autoencoder for detecting drift in industrial sensor data-based prediction.
- Welcome to Dr Jie Zhang, who joins us from Greenwich University to work on the Supply Chain Digital Twin project.
- Edward discussed his research at the Indonesian Current Affair Discussion held in Lady Margaret Hall, Oxford, exploring the philosophical idea of a virtual twin and how it is related to Digital Twins of today.
- Welcome to MAG 2020! After the Christmas break, we had a great start to the new year.
- Our new paper entitled “Extracting supply chain maps from news articles using deep neural networks”, has been accepted for publication in the International Journal of Production Research. The paper details Pascal Wichmann’s PhD research in the use of Natural Language Processing for automatically generating Supply Chain Maps from publicly available data. This work has led to the exciting spin-off called Versed.AI.
- Our new paper entitled “Uncertainty in Neural Networks: Approximately Bayesian Ensembling” will be presented by Tim Pearce at the upcoming 23rd International Conference on Artificial Intelligence and Statistics on June 3 - 5, 2020 Palermo, Sicily, Italy.
- The first Pitch in Manufacturing Day was held at the IfM on 11th December. With 25 participants from Cambridge, Sheffield, Newcastle and Oxford; we discussed the role of IoT in: Enabling new business models, Integrating data across digitally enabled Supply Chains, and novel analytics applications such as quality prediction, and automation of IP licensing payments. Discussions also centred on how IoT can be adopted by SMEs, and how IoT can be integrated with legacy systems.
- Our quarterly DO-TES meeting has been successfully completed at Ansys in Sheffield- we are pushing the boundaries of digital spare parts prediction and planning with use cases from Rolls Royce.
- Value Chain Lab has visited us to present their freight collaboration solution approaches and we discussed how agent-based systems could help with supply chain operations planning.
- We are partnering with Fetch.ai to develop use cases in Smart Supply Chains. Fetch.ai is a Cambridge based AI and blockchain company that is building an innovative platform for connecting IoT devices and algorithms to enable collective learning. MAG’s research in multi-agent-based supply chains are a perfect match to Fetch.ai’s vision of democratic, decentralised data and AI. Stay tuned to find out more!
- Dr. Luiz Fernando Bittencourt, from the University of Campinas (UNICAMP), Brazil, gave a seminar on “Cloud Computing, Fog Computing, and Internet of Things: Data Processing for Smart Environments in the Era of Industry 4.0” - resulting in exciting discussions around how trade-offs between the computational cost of communication, memory and speed can be addressed via distributed computational layers and how these new architectures may help enable smart manufacturing and supply chains.
- Dr. Griffith Rees from the Sheffield Quantitative Methods Institute gave a seminar on modelling the effect of spatial contagion on the adoption of the pre-internet FidoNet network. FidoNet began in 1983 and peaked in 1995 with users across six continents and up to 5 million users in the US alone. It persists to this day but declined by an estimated three orders of magnitude. The seminar generated discussions around contagion models, and how learnings from FidoNet can be relevant to today’s social networking websites.
- Farewell to Jose Luna Andrade - who has been with MAG as a visiting researcher in the past six months. Jose has been working on developing an analytical approach for the evolution of countries exports and how competitive advantage can be benchmarked. MAG wishes all the best to Jose!
- Thiago, Hang and Alexandra visited Rolls Royce for our DOTES project - as part of DOTES, we are in the process of creating prediction models for remaining useful life estimation which is in turn used for optimising the spare parts supply chain. Find out more about the project here.
- Liming demonstrated the first version of our Autonomous Supply Chain platform at the Advanced Engineering Show in Birmingham. At the same show, Alexandra gave a talk on how IoT based Data Analytics can be used to create value in Supply chains and Manufacturing.
- Our new paper “Supply chain data analytics for predicting supplier disruptions: A case study in complex asset manufacturing” is published in the International Journal of Production Research.
- Dr. Siddharth Arora from University of Oxford, gave a seminar on “From personalised medicine to decision-making for emergency care: Role of smart devices and analytics” giving interesting insights into how data-driven techniques can be used in predicting patient waiting times in emergency departments, which in turn can help improve overall patient satisfaction, and staff and resource allocation.
- Dr. Houyuan Jiang, from Judge Business School, gave a seminar on “Reconfiguration of Accident-and-Emergency Facilities in England”. His talk focussed on the use of queuing models for analysing cost and service implications of specialist and generalist facilities in hospital emergency departments.
- Xiang presented his work on “Multi-Agent System for Machine Learning Under Uncertainty in Cyber-Physical Manufacturing System” at the SOHOMA’19 International Workshop on Service-oriented, Holonic and Multi-Agent Manufacturing Systems for Industry of the Future. The theme of SOHOMA’19 Workshop is “Smart anything everywhere – The vertical and horizontal manufacturing integration”.
- Alexandra joined Camb.AI’s panel discussion on “Blurred Lines: The impact of AI in Business and Academia.” The discussions focussed on the role of regulation in ensuring transparency and ethical AI, and the challenges of integrating business-oriented and academic work in AI.
- Many congratulations to Dr Liming Xu for passing his PhD defence at the University of Nottingham!