The three-day Deep Learning IndabaX, a locally organised version of the annual Deep Learning Indaba, took place at UKZN’s UNITE Building was attended by researchers, students and experts in machine learning from across Africa.
The event aims to ensure an even spread of knowledge sharing and capacity building in machine learning across the continent. It enables experimentation with methods to strengthen African machine learning and its local communities; allowing more people to contribute to the conversation around machine learning and artificial intelligence (AI).
Lecturer in Computer Science at UKZN, Mr Anban Pillay; postdoctoral researcher, Dr Maria Schuld from the Centre for Quantum Technology; and Physics Master’s candidate, Ms Amira Abbas, joined the team of organisers which included representatives from the University of the Witwatersrand, Wolfram Research, Numberboost, and Nedbank.
The first day began with a discussion on AI in Africa. Panellists included Professor Tshilidzi Marwala, University of Johannesburg Vice-Chancellor; UKZN’s Professor Francesco Petruccione; Dr Benjamin Rosman; UKZN alumnus, Dr Justine Nasejje; Ms Pelonomi Moiloa; Ms Jade Abbott; and Mr Sicelukwanda Zwane. The discussion highlighted AI issues in the technical sector and approaches to solving these through education and research. It emphasised the importance of collaborating – not only internationally, but also regionally in order to promote research-sharing among African technologists.
‘Events like these are a good example of opportunities for outreach initiatives,’ said Nasejje, who suggested that the South African government follow the example of other countries by funding high school learners’ attendance at such events to increase their exposure to the field.
During the panel discussion, Marwala suggested that technology students engage with subjects in Humanities and Sociology. ‘They shouldn’t be so embedded in technology that they forget the human element,’ he said.
Keynote presentations were made by Marwala on rationality of AI machines, and Oxford University postdoctoral researcher, Dr Andrew Saxe, on the topic of the high-dimensional dynamics of generalisation error in neural networks. ‘Professor Marwala’s talk motivated me to take note that pre-processing is actually good material and is not to be taken advantage of. It helps with AI insights and helps bring better solutions,’ said Ms Nakedi Letsoalo from Explore Data Science Academy.
‘It was really inspiring to see the scale of the interest and level of ideas people are discussing. I hope that these links between international and regional will grow strong and I hope to come back in a few years and see people building on these ideas,’ said Saxe.
Parallel sessions throughout the indaba featured a diverse range of speakers that covered foundational topics, research, industry, ethics and policy.
The third day of the event featured a Hackathon and an Unconference; with Hackathon sponsors setting the challenges and awarding winning teams.
‘The whole event was very informative, insightful and technical; I learnt so much that I will apply in my research,’ said University of Cape Town master’s student, Ms Zinzi Villo.
‘I gained substantial information about deep learning and its applications. It’s important as a researcher to be able to integrate different disciplines in order to improve current research methods, and this conference gave me many actionable take-aways that I could apply to my field,’ said UKZN PhD Chemistry graduate, Dr Adele Cheddie.
Words: Christine Cuénod
Photograph: Albert Hirasen