The proliferation of quality software implementations of sophisticated machine learning frameworks has enabled scientists from various fields to rapidly spin up powerful models for forecasting, classification and prediction tasks. We often find, however, that lack of understanding and rigor in model development can lead to sub-optimal results and results that are not readily reproducible. In this work package we aim to provide a one-stop collaboration shop for scientists seeking help in the implementation of ML tools for their research. 

Participants

• PI: Stefan Lotz (South African National Space Agency, North-West University)

• Tian Theunissen (MUST Deep Learning, North-West University)

• Marelie Davel (MUST Deep Learning, North-West University)

• Randle Rabe (MUST Deep Learning, North-West University)

• NITheCS Associates

Objectives

To foster and maintain lasting collaboration with other NITheCS researchers seeking to apply deep learning techniques to their research problems.

Activities and Methodology

In collaborating directly with scientists across various fields we aim to position this work package to nurture collaboration and skills development in the application and utilisation of ML methods. To foster collaboration and enhance the deep learning skills of South African researchers we will establish standard operating procedures for the application of deep learning to different problem classes. This work package is complementary to work packages 1 and 2 since these work packages aim to foster collaboration.

Participants

• PI: Inger Fabris-Rotelli (University of Pretoria, NITheCS Associate) 

• Justin Harvey (Stellenbosch University, NITheCS Associate)

• Bruce Bassett

• Priyanka Nagar (Stellenbosch University, NITheCS Associate)

• Ansie Smit (University of Pretoria)

• Sonali Das (University of Pretoria)

• Daniel Maposa (University of Limpopo)

• Danielle Roberts (University of Kwazulu-Natal)

• Fabio Correa (University of the Free State)

• Michael von Maltitz ((University of the Free State)

• Bruce Bassett (WITS and UCT)

• StatSNetSA members in Statistics across South Africa from every institution

• NITheCS Associates

Objectives

To foster and maintain lasting collaboration with other NITheCS researchers seeking to apply Statistics  and Advanced Analytics to their research problems.

Activities and Methodology

In collaborating directly with scientists across various fields we aim to position this work package to nurture collaboration and skills development in the application and utilisation of Statistics. A specific need is the upskilling in Statistics in postgraduate studies for graduates who haven’t majored in Statistics. This leaves a gap in further growth. The idea of a postgraduate intern will be investigated. This will be like a research associate sitting in a Statistics department collaborating on Statistics research without registering for a degree, to enable growth and/or to enable RPL into postgraduate studies in Statistics and Data Science.