BEGIN:VCALENDAR
VERSION:2.0
PRODID:talks.ox.ac.uk
BEGIN:VEVENT
SUMMARY:Unbiased Bayes for Big Data: Paths of Partial Posteriors - Heiko S
trathmann (Gatsby UCL)
DTSTART;VALUE=DATE-TIME:20150225T130000Z
DTEND;VALUE=DATE-TIME:20150225T140000Z
UID:https://talks.ox.ac.uk/talks/id/f5480b0d-e9f9-4410-a5ec-bf97c29feb9e/
DESCRIPTION:*Abstract*\n\nA key quantity of interest in Bayesian inference
are expectations of functions with respect to the posterior. Markov Chain
Monte Carlo is a fundamental tool to consistently compute these expectati
ons via averaging samples drawn from an approximate posterior. However\, i
ts feasibility is being challenged in the era of so called Big Data as all
data needs to be processed in every iteration. Realising that such simula
tion is an unnecessarily hard problem if the goal is estimation\, we const
ruct a computationally scalable methodology that allows unbiased estimatio
n of the required expectations without explicit simulation from the full p
osterior. The scheme's variance is finite by construction and straightforw
ard to control\, leading to algorithms that are provably unbiased and natu
rally arrive at a desired error tolerance. This is achieved at an average
computational complexity that is sub-linear in the size of the dataset. We
demonstrate the utility and generality of the methodology on a range of c
ommon statistical models applied to large scale benchmark and real-world d
atasets.\n\n--\n\n*Speaker's bio*\n\nHeiko Strathmann first studied Jazz g
uitar in the Netherlands\, then did a BSc in Computer Science in Germany\,
followed by a MSc in Machine Learning at University College London. Since
2013\, he is a PhD student at the Gatsby Unit for Computational Neuroscie
nce and Machine Learning at UCL. He is an open-source activist and one of
the main developers and organisers of the Shogun Machine Learning Toolbox.
\nSpeakers:\nHeiko Strathmann (Gatsby UCL)
LOCATION:Department of Statistics (Common Room)\, 24-29 St Giles' OX1 3LB
TZID:Europe/London
URL:https://talks.ox.ac.uk/talks/id/f5480b0d-e9f9-4410-a5ec-bf97c29feb9e/
BEGIN:VALARM
ACTION:display
DESCRIPTION:Talk:Unbiased Bayes for Big Data: Paths of Partial Posteriors
- Heiko Strathmann (Gatsby UCL)
TRIGGER:-PT1H
END:VALARM
END:VEVENT
END:VCALENDAR