Max Mowbray

Our review paper on the use of Machine Learning in the chemical and process industries is now available in Reaction Chemistry & Engineering. If interested, the paper is freely available here: lnkd.in/eJnWNk6f

Is there not something slightly immoral about large energy/fossil fuel firms engaging in share buyback programs instead of reinvesting in the energy transition?

Maybe all publications

The more I review, the more I think double blind is a better idea for journal publications

@RestIsPolitics would be very interested to hear your opinion on @MazzucatoM’s work at UCL and how/if it fits into your experience within Government. Perhaps even a leading episode?

Had a great time presenting, catching up and meeting new people at #escape33 in Athens this week. Looking forward to Florence next year!

Editor's Choice: Integrating process design and control using reinforcement learning @AntonioE89 @Pnosapp @MaxMowbray3 @StevenSachio1 @maria_papathan @ImperialChemEng @sargent_centre @uclchemeng @ChemEngManUni #ChERD doi.org/10.1016/j.cher…

Great to hear about @Pnosapp’s journey from Athens to Manchester to London and then to Cambridge today at @OfficialUoM. Thanks for the great talk @Pnosapp !

Are you interested in outreach? The local section has 2 opportunities in March. Please get in touch for more information. #outreach #chemistry #stem @RoySocChem @rscecnmanc

What’s Broken with RL Research and a Potential Fix: open.substack.com/pub/avivtamar/…

Thanks to Miguel and Guoquan for the collaboration!

Additionally, we propose risk-sensitive formulations that protect against low probability, high severity scenarios.

Our work on Reinforcement Learning (RL) for inventory management in multi-echelon supply chains is out in Digital Chemical Engineering. If interested please find the paper at lnkd.in/ebB5R56b

@EugeneVinitsky It could be any of these… and more! There are also lots of interesting problems beyond closed loop optimality: handling operational constraints, distribution shift, notions of robustness + stability, defining real systems as MDP, practical benchmarks that aren’t Atari, etc!

DP gets 100% but RL won’t on the same tasks. What’s going awry? Function approximation? Value overestimation? Bad choice of action parametrization?

Opportunity for research: on many tasks RL flatlines slightly below 100% i.e. 99.8%, it doesn’t quite have the robustness for a real world tool or product. What’s missing? How do we get perfect performance?

'There is NO cutting back. There is NO Money Saving Expert. You could put me into one of those households and do every trick in the book and I wouldn't even get close to scratching the sides of what is needed.' @MartinSLewis says how he'd stop rising energy bills. @mrjamesob

First day of @bluedotfestival underway!! Visit us in Tranquility Base for hydrogen-fuelled fun and 3D-printed figures from recycled coffee cups ☕️

@Pnosapp @UoMSciEng @cafe4dm @DongdaZhang Thanks Panos!

@MaxMowbray3 @UoMSciEng @cafe4dm @DongdaZhang Great work as always @MaxMowbray3!

Our paper on soft-sensing of end product quality in batch process systems is now out in Chemometrics and Intelligent Laboratory Systems. Work conducted @UoMSciEng @cafe4dm with @DongdaZhang. If interested, please find the open access link @ doi.org/10.1016/j.chem…

Last chance to join us at this year’s Bluedot festival! Help us enthuse the public about all things science and sustainability during the day. In the evening, you’re free to enjoy the rest of the festival! Send us a message or reply to the email below 👇

Now published - Integrating process design and control using reinforcement learning @AntonioE89 @Pnosapp @MaxMowbray3 @StevenSachio1 @maria_papathan @ImperialChemEng @sargent_centre @uclchemeng @ChemEngManUni #ChERD doi.org/10.1016/j.cher…