Hosted by UKRMP Smart Materials Hub, Imperial College Machine Learning Initiative and the Quantitative Sciences Research Institute

Why are we running this workshop?

The area of Regenerative Medicine, in which materials can be engineered to replace or repair damaged tissues, is a rapidly growing and exciting field and holds promise to revolutionise medical treatment. Experimentalists spend a lot of time optimizing numerous conditions: what temperature? which buffer? what pH? Classically, they hold all variables constant except one, vary that, decide on the best, and then lather, rinse, repeat with the rest. But what if we didn’t pick the best value at the beginning? Perhaps that initial decision meant we never actually found the best overall set of conditions. How do we know that we explored a wide enough range with enough replicates to make a well-informed decision? These are the questions that the statistical field of Design of Experiments seeks to answer.

Who should attend?

Both experimentalists that want to collaborate with machine learning experts to improve their experiments AND statisticians, mathematicians and other methodologists who want to learn about interesting problems faced by applied scientists and see how their methods can help!

The UK Regenerative Medicine Platform Smart Materials Hub led by Prof Molly Stevens has teamed up with statisticians Drs Sarah Filippi and Seth Flaxman from the Imperial College Machine Leaning initiative to host this workshop to bring mathematicians and statisticians together with experimentalists to explore how we can apply the right mathematical /statistical methods to the right experimental questions. We will hear from experts in the area of experimental design and then move onto case studies in how the field has helped and can continue to help benchtop scientists get the most information about their systems with the fewest number of experiments. This will also be an opportunity for methodologists to learn about interesting problems faced by applied scientists.

Bayesian optimization and active learning have become widely adopted in drug discovery to identify potential therapeutics and realise models for identification of the activities of specific subsets of compounds. Active learning describes a setting in which initially little or no data is available, and over time a learning algorithm is able to decide which data to acquire by addressing the so-called exploration/exploitation tradeoff. In this workshop, experts in machine learning and (Bayesian) multi-objective optimization will explain state-of-the-art approaches for experimental design.

This workshop provides an excellent opportunity to enhance and generate new partnerships between theoreticians working on experimental design and experimentalists alike.

Find out more and register

For up-to-date information about this workshop, and to register, please see the workshop website.