Is this course right for me?
Target Audience: Academics/Researchers/KE Staff
We invite you to attend an in-person workshop on ML/AI on Thursday the 18th of May 1.05 – 2.55 in McCance MC301. This third workshop is part of a series of workshops exploring ML/AI opportunities for collaboration with University colleagues. Prior knowledge of the topic is not required.
1.05-1:10 Welcome Professor Sergey Kitaev (Faculty of Science, Associate Dean for Research)
1:10-1:40 Dr Paul Murray (EEE) : AI and machine learning for image and video processing
1.40-1.55 Dr Olga Bylya (AFRC, Digital Factory): Investigation of the applicability of Artificial Intelligence models for microstructural prediction during hot forging of Inconel 718
1.55-2:10 Dr Clemens Kupke (CIS): Logics for AI
2:10-2:40 Dr Daniel Markle (SIPBS): ML-Assisted Accelerated Design and Development of Drug Products
2.40-2.55 Round up and discussion
The abstracts for the talks are available below.
Should you have any questions about the workshop, please contact Associate Dean Research Professor Sergey Kitaev at email@example.com.
Speaker: Dr Paul Murray
Title: AI and machine learning for image and video processing
Abstract:For the last decade, image and video processing research has been dominated by AI and machine learning. Deep learning is of particular interest to the community and solutions for image classification, object detection and image segmentation are wide ranging. In this presentation, Dr Murray will describe some of the research activity of his group which aims to develop machine learning and AI based tools for image and video processing to address a number of practical, industrial challenges.
Speaker: Dr Clemens Kupke
Title: Logics for AI
Abstract: In this talk I'll discuss close links of my research in computational logic with the areas of model learning and knowledge representation.
Speaker: Dr Daniel Markle
Title: ML-Assisted Accelerated Design and Development of Drug Products
Abstract: Traditional methods of developing drug products for a new active pharmaceutical ingredient are time-consuming, costly and often inflexible. The selection of the right excipients in tablets and process conditions are crucially important as they can impact manufacturability, performance and stability of the drug product. Formulation optimisation studies are conducted to identify a robust formulation that can meet manufacturability criteria (e.g. flowability, tensile strength) while fulfilling the desired performance targets, e.g. release of > 80% of the drug in less than 30 min. This is a multidimensional problem with a high degree of interdependence between raw material attributes, process parameters, and drug product properties. These complex relationships cannot be fully captured by first principle models and it is not feasible, in a reasonable time, to experimentally investigate these multidimensional formulation (type of excipient, concentration, drug loading) and process parameter (e.g. compression force, dwell time) spaces following traditional experimental planning and methods. This talk will present a hybrid system of models, including mechanistic and data-driven approaches, to predict critical powder blend (e.g. flowability) and quality attributes (tensile strength, porosity) of pharmaceutical tablets from raw material properties. I will further discuss how the hybrid modelling approach will enable a model-based optimisation of formulation and process conditions through our a high-throughput data-intensive micro-scale tablet development system that can automatically prepare and measure powder, and produce and test single tablets. The proposed system holds the potential to significantly accelerate the development of drug products while reducing costs and waste.
Delivered By: Faculty of Science and Faculty of Engineering