98abe15cda69635ba6e8c2f13bbf0c9a.ppt
- Количество слайдов: 13
Why is a new type of mechanism needed? • Although some UK universities produce world leading research • The uptake of this by industry is generally poor • There is a difference between curiosity-driven research • And the research needed by industry • The timescales are also different • Industry wants results in months not years! www. acrc. com 2
A new way • The ACRC will engage directly with industry • Carrying out research and development projects to solve problems specified by industry • The ACRC’s teams of researcher developers will deliver results on industry timescales to industry quality • Commercial developments will also be carried out • We need to develop a regular income stream through commercial and grant funding www. acrc. com 3
Core themes • Big data – data and text analytics – building business intelligence from disparate sources • Large-scale simulation and control of complex systems – predicting social/economic behaviour in critical environments – e. g. transport infrastructure, complex organisations such as hospital A and E departments • Software testing and quality assurance – ensuring software, apps and devices are fit for purpose – device library • Funding model – industrial subscriptions, R&D contracts, EU and UK (eg. TSB) programmes www. acrc. com 4
• • • • Testing and QA Automated testing using a variety of techniques Model-based testing - testing what you want Reverse Engineering ‘Discovery’ testing – building models of legacy systems -- recovering the modular structure of an application User testing – understanding the organizational context of systems App testing – on multiple platforms (device library) Evolutionary testing of large codes Automated code coverage -- C, Java Automated worst case performance testing: -- speed -- memory consumption -- energy usage www. acrc. com 5
Text analytics • identification of connections between key entities • topic modelling • document similarity • sentiment analysis • summarisation of multiple documents • knowledge discovery • machine translation Information Retrieval • identification of interesting documents from large collections (in both mono-lingual and multi-lingual settings) • exploratory search - alternatives to standard keyword-based search) • visualisation of document collections • recommendation (providing recommendations of interesting content from a collection) 6 • analysis of social media (e. g. Facebook, twitter)
Very large-scale agent-based modelling of complex systems, and complex socio-technical systems • Existing models and capabilities: • Social and economic systems • Crowd behaviours (social and cognitive) in detailed architectural settings • Traffic simulation • Biological systems (from molecular, cellular and tissue levels to sperm) • Platforms • HPC simulation, • GPU simulation, • Saa. S (simulation as a service) 7
Agent-based modelling with HPC - biology The Blood Brain Barrier is a mayor obstacle to effective neurological disease treatment Simulation of blood flow through capillaries in the brain studying the behaviour of nano particles and key receptors and gap junctions 8
Agent-based modelling with HPC - 2 • EURACE - the largest, most detailed and complete model of the European economy • Integrates all the different markets together - labour, credit, financial, goods etc. • Uses large numbers of complex agents - consumers/households; firms, banks, etc. • Demonstrates much more accurate simulations than traditional ‘equilibrium’ models • Used for policy modelling • eg. explains why QE does not lead to freer credit, etc. • Invited presentations to: • Bank of England, • HM Treasury 9
Agent-based modelling with GPU Pedestrian modelling in an airport context P. Richmond et al 10 www. acrc. com
Other applications of FLAME • Investigation of oxygen respiration in E. coli (Green et al, PLOS Comp. Bio. (2014) • Role of TILLR and actin in innate immune system (Qwarnstrom et al) • MAPKinase cross talk and compartmentalisation effects (Kiss-Toth et al) • Wound healing in epithelial cells (Walker et al) • Foraging in social insects (Ratnieks et al) • Social capital networks (Prell et al) • Coevolutionary marketing strategies in retail economies (Kiran) • Managing complex organisations (current commercial project) • Management information system for dynamic public spaces (current commercial project) 11
New types of Management Decision Support Systems • Complex systems involving people can lead to problems of overcrowding and overload • Transport interchanges when overcrowded need effective interventions to relieve the load • Systems such as hospital A and E can suffer large peaks in demand • Detailed simulations can provide information about what is the best sort of intervention • Software is being developed that will run multiple FLAME simulations faster than real time to provide best advice on which intervention to adopt under the precise current circumstances • Transport interchanges need FLAME. gpu • Hospital systems use FLAME HPC
Some of our supporters