UP Graduate’s Thesis Useful For Disease Diagnosis, Spam Detection

University of Pretoria (UP) graduate Mia Gerber recently received a master’s in Computer Science with distinction for her thesis investigating automating the design of the deep neural network pipeline to address fourth industrial revolution (4IR) problems by using artificial intelligence.

Gerber completed her degree in a year and a half following her return to academia after a work stint.

The machine-learning techniques that she developed as part of her research not only makes artificial intelligence more accessible but can also be applied to areas that are relevant to the United Nations’ Sustainable Development Goals (SDGs): disease diagnosis, spam detection and sentiment analysis.

In 2018, Gerber obtained a BSc in Computer Science (Hons) from UP, with distinction, before going on to work at an insurtech start-up and returning to UP to further her studies.

She has been accepted for a Ph.D. at the University and plans to expand on her master’s research.

Gerber’s thesis investigated automating the design of the deep neural network pipeline to address fourth industrial revolution (4IR) problems by using artificial intelligence.

Deep neural networks are machine-learning techniques that mimic the way the brain works.

When applying these networks to a new application, the design pipeline involves several tasks, such as determining the most suitable neural network architecture; this is time-consuming and requires expert knowledge.

Gerber’s research has contributed to the body of ongoing research into automated machine-learning and will facilitate the design of machine-learning techniques to enable non-experts to apply them to solve real-world problems.

“The process of doing my master’s degree has been a learning experience like no other,” Gerber says.

“I am blessed to be able to continue my studies and to wake up every day and do what I love.

“I am thankful for the continual support of my supervisor and research group.”

Gerber, who is a research assistant in the Nature-Inspired Computing Optimization Group (NICOG) research collective, says she was able to focus fully on her studies because “she loved what she was doing, and particularly because it would have a positive impact on society”.

The research group aims to address intractable challenges by using machine-learning and optimisation techniques that take analogies from nature – such as genetic algorithms, which is a way of solving optimisation problems based on a natural selection process that mirrors biological evolution.

“Gerber is passionate and has a strong aptitude for her chosen field of study,” says her supervisor, Prof Nelishia Pillay, who holds the National Research Foundation (NRF)/Department of Science and Innovation (DST) SARChI Chair in Artificial Intelligence for Sustainable Development and the Multichoice Joint Chair in Machine Learning.

“The NICOG’s work is tied to the 4IR and the SDGs; these issues were covered by Gerber in terms of making artificial intelligence accessible,” said Prof Pillay.

“I’m pleased to note that her work is extending towards a Ph.D., she has also completed a draft for her first journal paper.”

Prof Pillay adds that Gerber contributes significantly to the research group and that she was pleased that her academic work was aligned with the betterment of society in line with the SDGs.

“To predict diseases, you need neural networks, which are used for forecasting,” Prof Pillay explains.

“Deep neural networks have been able to solve several 4IR problems. No matter which industry you’re looking at, it’s about finding the solution.”

Gerber says: “The science has already been done”.

“My work is largely about making artificial intelligence more accessible.”