AI/ML in Regulated (GxP) Life Sciences Sectors - Concept/Project Phase: Representative Data Sets
Complimentary
Learning Level: Basic
Time: 1000 - 1100 EDT
Session Length: 1 hour
Guiding the pharmaceutical and medical device industries to successfully transition into the eras of AI/ML and Pharma 4.0, the presenters will provide tangible insights derived from GAMP and quality perspectives to ensure product quality, patient safety and data integrity. Following the first webinar of the series on the concept stage, we will concentrate on the fuel that drives any Machine Learning system: data. Covering conceptual aspects such as types of data and how they are linked to potential model types, we will elaborate on data quality and representativeness considerations as key preconditions for powerful Machine Learning systems that fulfill their intended use. Finally, we will provide guidance on practical experience on data splitting strategies and potential challenges to ensure representativeness throughout the ML subsystem life cycle. Example studies and use cases will accompany this webinar to provide further practical insights.
Learning Objectives
- Provide an overview of applicable types of data and potential use of multi-modal data as input to machine learning systems. Describe data preparation and pre-processing methods, including the labeling, augmenting, and homogenization of data for machine learning models.
- Deepen the understanding on the relevance of data quality and representativeness for the expected performance of the model in the real-world setting. Particular risk implications on these topics will expand on the Risk and Control Framework presented in the first webinar, while data governance principles will be presented to facilitate trustworthy and high quality data.
- Provide guidance on data splitting strategies to arrive at suitable training, validation and test data sets that suitably balance requirements for training and quality assurance.
