IST-Africa 2023, the eighteenth edition of the Ministerial Level Technology Research Conferences, is held in Tshwane, South Africa, from 31 May to 02 June 2023. The event is hosted by the Government of South Africa through the Department of Science and Innovation, with the support from the European Commission (EC) and the African Union Commission (AUC). IST-Africa Week 2023 provides a prestigious global platform to showcase existing technology-enabled research, innovation, and ICT4D initiatives and capabilities across Africa, Europe, and other regions. The Conference Programme encompasses various components, including technical and policy papers, case studies, and workshops. The primary objectives of the IST-Africa Conference Series are to facilitate the development of the Information Society and Knowledge Economy in Africa, foster international research innovation, promote policy cooperation and coordination, and encourage the utilization of research outcomes by both the public and private sectors.
Vianney Kambale's Research Addresses the Need for Transfer Learning Guidelines in Time Series Forecasting
At this conference, Vianney Kambale presentes a research paper titled “Sensitivity Analysis for an Ensemble Transfer Learning Scheme for Time Series Forecasting – Case-study of a Shallow Neural Network Architecture”. In his presentation, he talkes of the fact that Transfer Learning (TL) has emerged as a significant topic in machine learning, particularly in the context of time-series forecasting. His research paper aimed to address the existing gap identified in recent surveys, which highlighted the need for empirical studies to develop guidelines for transfer learning approaches and method design selections that can benefit practitioners. The paper proposed a comprehensive sensitivity analysis methodology for TL schemes, specifically tailored to a given machine learning model.
The methodology begins by suggesting and providing comprehensive definitions for five key TL performance metrics. Subsequently, the core steps of the sensitivity analysis process related to TL are formulated. To illustrate the approach, a mini-sensitivity analysis was conducted on a shallow MLP (Multilayer Perceptron) network. Despite its smaller scale, the mini-sensitivity analysis validates the applicability of the proposed methodology and offers intriguing insights. Ultimately, the outcome of this project has the potential to lead to the development of a pre-trained model that can greatly assist time-series forecasting practitioners.
Vianney Kambale's PhD Research at Klagenfurt University
Vianney Kambale is a recipient of the OeAD scholarship and is currently conducting his PhD research on the topic “Robust Time series Forecasting involving Advanced Deep Learning Concepts and Causal Methods” at the Klagenfurt University in Carinthia, under the supervision of professor Kyandoghere Kyamakya. His research interests include the application of Machine Learning, Deep Learning and AI techniques in the analysis and forecasting of time series. Moreover, Vianney Kambale is passionate about research, teaching and learning with a genuine desire to make a positive impact on individuals and society through education.
Join us in celebrating Vianney Kambale's remarkable contribution to the field of time series forecasting at the IST-Africa Conference. His research on sensitivity analysis for transfer learning provides valuable insights for practitioners. Stay tuned for more groundbreaking developments from this talented researcher.