skip to primary navigationskip to content

CEB Machine Learning Workshop

last modified Jun 21, 2018 01:52 PM

On 7 June 2018 the Sustainable Reaction Engineering group (led by Prof. Alexei Lapkin) ran a one-day open workshop on the application of advanced statistics and machine learning in (bio)chemical engineering. This workshop was the first impact event of the new £1.2m EPSRC project "Combining Chemical Robotics and Statistical Methods to Discover Complex Functional Products", led by Prof. Lapkin at CEB in collaboration with Prof. Leroy Cronin at the University of Glasgow (Chemistry) and Prof. David Woods at the University of Southampton (Statistics).

The workshop was attended by over 50 machine learning enthusiasts, including researchers and academics from CEB, the Cambridge Computer Laboratory, Cavendish Laboratory and the Engineering and Biochemistry Departments, along with a number of visiting research students and those from industry. The workshop covered such topics as the discovery of new chemical reactivity, training of predictive reactivity models on chemical data, selection of optimal surrogate models and non-parametric multi-objective optimisation. Dr Alpha Lee (Cavendish) and PhD student John Bradshow (Engineering, working in the group of Dr Miguel Lobato) presented two different approaches to predicting chemical reactivity by machine learning models. This topic was complemented by Dr Philipp-Maximilian Jacob’s (CEB) talk on statistics of chemical data and statistical prediction of unknown reactions between known chemical species. A significant activity in machine learning within CEB takes place at Cambridge’s Singapore outpost – the Cambridge Centre for Advanced Research and Education in Singapore (CARES). Prof. Markus Kraft, Director of CARES and Professor of Chemical Engineering at CEB, gave a talk over Skype on the optimisation of surrogate functions selection.

The afternoon was dedicated to optimal design. PhD student Liwei Cao (CEB, SRE group) talked about developing optimal machine learning pipelines, which allow for increased model accuracy. The main part of the workshop was a seminar by Prof. David Woods (Southampton) on developing sequential optimisation methodology for chemical reactions. This work has paved the way for several new academic and industrial projects.

Two such projects were presented at the workshop by their lead academics, including Prof. Lapkin's "Combining Chemical Robotics and Statistical Methods to Discover Complex Functional Products". Dr Richard Bourne (University of Leeds) presented the four-year project "Cognitive Chemical Manufacturing", also funded by the EPSRC. Both projects are funded by the 'Manufacturing the Future' theme of EPSRC and fall within the remit of the EPSRC Grand Challenge Network 'Dial-a-Molecule'.

The very last session of the workshop turned into a collaborative project meeting where members of the two projects’ teams discussed areas of synergy and the potential for sharing equipment and software codes, having some shared chemical examples. They also agreed on the need for joint work on the standardisation of the symbolic interface from chemical reactions to instrument control.

If you want to learn more about the ongoing work in CEB in chemical robotics, do contact Alexei Lapkin on .