The 2018 Deep Learning Indaba in September was a week-long gathering of more than 500 students, professors, industrial researchers and other AI practitioners, almost exclusively working on the African continent.
Its goal was to share and celebrate ongoing research in Africa around artificial intelligence. I ‘d been asked to give a lecture there on convolutional networks (CNNs) and was quick to accept given the uniformly positive feedback I ‘d heard about the 2017 edition. I wasn’t disappointed.
The deep engagement of the students throughout the week left a lasting impression on me. During the poster sessions, I had the opportunity to interact with many students and learn about the dazzling variety of applications that they were addressing. Among my favourites were the work of Raesetje Sefala and colleagues on using satellite imagery to predict spatial apartheid via demographic and socioeconomic data, and work from Luyolo Magangane and colleagues on inferring relations between entities using recursive neural tensor networks.
The degree of interaction during (and after) my lecture really marked me. I began with an introduction to discrete convolution and cross-correlation before describing convolutional variants and deep convolutional networks, and ending with a sample of application areas and related models. The attendees were quick to ask for clarifications or to dig deeper on points which made for a really rich session (certainly for me).
The Indaba also presented a unique opportunity for attendees to listen to world-renowned experts in a variety of areas including reinforcement learning, recurrent networks, production-ready deep learning systems and more. In spite of there being more than 500 people it was fairly easy for students to interact with lecturers and I personally had many interesting conversations with students eager to discuss their work and brainstorm ideas for tackling their current research challenges. I also learned a lot about fields of application from radio astronomy to plant pathologies. In addition to keynotes and lectures, there were also interesting breakout sessions covering topics from “AI for Africa” to “Generative Models and Healthcare” and “Life of a Machine Learning Startup”.
I left the Indaba with the impression that the Indaba community had unbridled enthusiasm and a deep sense of purpose captured in the theme for the Indaba which was “Masakhane!” meaning 'we build together'.
I’d like to end by both congratulating and thanking the organizers for their hard work which was instrumental to the Indaba’s success. Best of luck for next year’s edition in Kenya!
Material from the DLIndaba are available online.
Naila Murray is Senior Scientist and Group Lead of the computer vision research team at NAVER LABS Europe.