BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Chemical Engineering and Biotechnology occasional
seminars
SUMMARY:Statistics and Machine Learning in (Bio) Chemical
Engineering - An Open Workshop - Prof Alexei Lapki
n et al
DTSTART;TZID=Europe/London:20180607T093000
DTEND;TZID=Europe/London:20180607T170000
UID:TALK106240AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/106240
DESCRIPTION:9:30 Introduction to the Challenge-led Talks/Discu
ssion (A. Lapkin)\n\nTalks and discussion on the t
opics of surrogate functions\, optimisation\, clas
sification\, data mining\, automation and more in-
depth topics\, such as specific methods of tacklin
g various uncertainties in model development.\n\n“
Interpretable ML for Chemistry: Designing algorith
ms with statistical physics and extracting chemica
l knowledge from results”\, Dr Alpha Lee\, Cavendi
sh Laboratory.\n\n13:00 Lunch break\n\n14:00 Semin
ar talk: “Closed-loop automatic experimentation fo
r optimisation”\nDave Woods\, Professor of Statist
ics in the Southampton Statistical Sciences Resear
ch Institute\n\nAbstract: Automated experimental s
ystems\, involving minimal human intervention\, ar
e becoming more popular and common\, providing eco
nomical and fast data collection. We discuss some
statistical issues around the design of experiment
s and data modelling for such systems. Our applica
tion is to “closed-loop” optimisation of chemical
processes\, where automation of reaction synthesis
\, chemical analysis and statistical design and mo
delling increases lab efficiency and allows 24/7 u
se of equipment.\nOur approach uses nonparametric
regression modelling\, specifically Gaussian proce
ss regression\, to allow flexible and robust model
ling of potentially complex relationships between
reaction conditions and measured responses. A Baye
sian approach is adopted to uncertainty quantifica
tion\, facilitated through computationally efficie
nt Sequential Monte Carlo algorithms for the appro
ximation of the posterior predictive distribution.
We propose a new criterion\, Expected Gain in Uti
lity (EGU)\, for optimisation of a noisy response
via fully-sequential design of experiments\, and w
e compare the performance of EGU to extensions of
the Expected Improvement criterion\, which is popu
lar for optimisation of deterministic functions. W
e also show how the modelling and design can be ad
apted to identify\, and then down-weight\, potenti
ally outlying observations to obtain a more robust
analysis.\n\n15:00 Introduction to the EPSRC proj
ect “Combining Chemical Robotics and Statistical M
ethods to Discover Complex Functional Products”\,
a collaboration between Universities of Cambridge\
, Glasgow and Southampton\n15:30 Coffee/Networking
\n16:00 Afternoon workshop session\n17:00 Close of
workshop\n\nPlease register at\nhttps://www.event
brite.co.uk/e/statistics-and-machine-learning-in-b
io-chemical-engineering-tickets-46075627442
LOCATION:Maxwell Centre\, Cavendish Laboratory\, JJ Thomson
Ave\, Cambridge CB3 0HE. Small Lecture Theatre
CONTACT:Alexei Lapkin
END:VEVENT
END:VCALENDAR