COOKIES: By using this website you agree that we can place Google Analytics Cookies on your device for performance monitoring. |

University of Cambridge > Talks.cam > Computer Laboratory Systems Research Group Seminar > Analysis of metrics for large graphs via approximate Bayesian computation with application in higher order graphs

## Analysis of metrics for large graphs via approximate Bayesian computation with application in higher order graphsAdd to your list(s) Download to your calendar using vCal - Damien Fay (University College Cork)
- Friday 14 December 2012, 16:00-17:00
- FW26, Computer Laboratory, William Gates Builiding.
If you have a question about this talk, please contact Eiko Yoneki. This talk has been canceled/deleted Approximate Bayesian Computation (ABC) is a parameter estimation technique that has become popular recently due to it applicability to situations in which direct estimation of the likelihood of parameters given the data is not possible or problematic. My talk will introduce some new material on the application of ABC to synthetic graphs and models for thermal flow in buildings. In particular ABC provides an estimate of the posterior probability of the parameters given a target. The first part of the talk will detail ABC (partly given in tutorial style for those interested in trying this themselves). I shall then detail how this can be applied in building energy models together with particle filtering. The second part will start with a comparison of 6 graph metrics via ABC in the case when the target graph is already known. In particular we show that one of the metrics is seriously flawed while the others demonstrate a trade-off between bias and variance in the parameter estimates. The first contribution is the method itself which can be applied to any graph topology generator. In the second part of the talk we present results on a target graph, the underlying mechanism of which, is not known; the full yeast graph is used as an example. We show that the BA topology generator* cannot generate graphs close to this target but still the ABC algorithm identifies those parameters which are closest (wrt a given metric). Much of the material in this talk is novel and represents ongoing research and so for us the aim is to get feedback and suggestions for future work. - Its normally called the AB generator but here we opt to swap the names to avoid confusion with ABC .
Damien Fay obtained a B.Eng from University College Dublin (1995), an MEng (1997) and PhD (2003) from Dublin City University and worked as a statistics lecturer at the National University of Ireland (2003-2007) before joining the NetOS group, Computer Laboratory, Cambridge from 2007 and again in 2010. He is currently a research fellow in UCC , Ireland. His skills lie in the application of statistics to computer networks, social networks and energy systems. His research interests include applied graph theory, time series analysis and statistical modelling, This talk is part of the Computer Laboratory Systems Research Group Seminar series. ## This talk is included in these lists:This talk is not included in any other list Note that ex-directory lists are not shown. |
## Other listsCambridge Carbon Footprint Cambridge University Self-Build Society Sir Richard Stone Annual Lecture## Other talksArriva Trains Wales by Tom Joyner Joinings of higher rank diagonalizable actions New micro-machines, new materials Seminar – The Cambridge Sustainable Food Hub Neurodevelopment disorders of genetic origin – what can we learn? Numerical solution of the radiative transfer equation with a posteriori error bounds |