Fully Funded Studentship – StatML CDT – Self-optimizing Experiments
To be eligible for this studentship, you need to be classified as a Home or EU student.
Deadline for Applications: 4 August 2019
Academic Supervisors: Professor Axel Gandy, Dr Nikolas Kantas (Imperial) – contact: firstname.lastname@example.org
Artificial intelligence (AI) systems comprise of modules that observe, learn, decide, and act. The goal of this project is to focus on the decisions that an AI system must take to run experiments autonomously in a chemical lab. The current project is part of a bigger research collaborations between Imperial and BASF.
The project aims to formulate a new paradigm for designing experiments using AI systems. This means we will integrate AI modules with an experimental setup so that the experiment automatically configures itself, assesses results and then re-updates input configurations continuously for different environments and settings. The main application of interest is flow chemistry in micro-fluidics, where one is interested to achieve a high-quality mixture concentration under a variety of temperature settings, reactions, conversion rates. etc. The problem has many challenges:
- One is interested in optimising multiple objectives for different potential environments.
- The environments are non-linear, have uncertainty and are hard to predict or model accurately.
- There are a large number of outputs and inputs.
As a result, traditional methods based on control theory and model estimation, are not applicable here. The project will develop and integrate tools from modern computational Statistics, Bayesian Optimisation and Reinforcement learning. The aim to integrate these tools in an AI system that achieves high quality outputs while at the same time reducing operational costs from data collection requirements or time duration of experiments.
The performance of the system will be assessed both numerically as well as theoretically, considering interpretability, robustness and generalisability.
The project will be conducted in close collaboration with BASF researchers based in Ludwigshafen, Germany, particularly from the Statistics, Machine Learning, and Artificial Intelligence group.
To apply, you need to complete the following two steps:
1) Submit an application to Imperial:
If you have already submitted an application for a PhD at Imperial there is no need to submit a new application. Simply include your application number in your response.
If you have not applied yet, please follow the instructions for applying to a Mathematics Research (PhD) at:
You will need to first select “Postgraduate Research” and then the Doctoral Programme entitled “Mathematics Research (PhD)”.
Put “StatML, Prof A Gandy, Dr N Kantas” into the supervisor field.
(You will receive an application number at the end of the application process, called “Eip ID” or “CID”)
2) Send an e-mail to statml.io.admissions@email@example.com the subject “Application – Self-optimizing Experiments.