The Indaba is a learning opportunity, but also an opportunity to connect different actors in the continent around Machine Learning and related topics. This proposal took shape after hearing about the successful experiences from a sister summer school in a different part of the world to build research bridges across the borders. We propose therefore to mimic their experience, and get researchers from different countries to come up with research projects.
How would it work?
The Ideathon is a competition where candidates propose research projects. The only requirements about the project are:
- It needs to include a machine/deep learning component. This component does not need to be at the core of the project.
- It needs to include researchers from at least two African countries. International collaborators are welcomed as long as at least two members of the team are based in two different countries on the continent.
A jury then selects the best 3 projects. The winners will receive support from the Deep Learning Indaba to turn their idea into an actual project in two forms:
- A seed funding
- Mentorship from experienced researchers
- Monday 22/08: Official start
- Monday 22/08 to Thursday 25/08: Meet potential partners, form the idea and prepare a 3 minutes pitch. There won’t be any dedicated time to work on this, but there will be many networking opportunities for you to meet potential partners and prepare your idea and pitch.
- Wednesday 24/08: Mentorship to refine the project and the pitch.
- Thursday 25/08: Present the projects – 3 minutes per team, subject to change in function of the number of participants
- Friday 26/08: Project selection and announcement
The project can be anything, as long as it has a machine/deep learning component. It can be a pure theoretical research pushing the understanding of one particular topic that you are passionate about, as it can be a pure applied research taking recent successful advances into an application that you find particularly motivating, or anywhere in between. The most important requirement here is that the proposal is made by at least two people from at least two countries in Africa.
This competition is happening over 4 days. Given the short period of time, we are not expecting full research proposals. On Thursday evening, you will have 3 minutes to convince the jury that your idea is interesting to pursue and that it is realisable. You will have complete freedom on how to shape this presentation. Here is a non-exhaustive list of questions that you could use for inspiration:
- What do you propose to study and what is your motivation? What could be the impact of this research locally? For the ML community in general?
- How much would you expect such a research to cost? What are the needed resources for it? Do you need additional/external collaborations? Are there structures that could help you with this?
- What makes your idea realisable? What are the challenges you could face? What are the factors that could facilitate the realisation of this idea?
Your projects will be evaluated by a diverse jury. The judging will be through a standardised scoring form, focusing on three axes:
- Motivation and Impact: Is the idea original? Is it well motivated and sound?If conducted successfully, would it have a significant impact?
- Feasibility: Is the idea likely to lead to a successful project? Have the team considered potential challenges?
- Team: geographic, gender, language and other dimensions of diversity.
More details on the judging procedure will be shared at the start of the Indaba week.
The DLI support
We will support you during this new experiment in different ways, both during the indaba week and after the Indaba week.
During the Indaba week: On Wednesday afternoon, we will offer you mentorship to refine your project definition and pitches. As soon as you meet your potential partner, make sure to register your interest using this form by Tuesday EOD (Tunisia time). This would allow us to plan for the mentorship session and for the presentations accordingly, and to match you as much as possible with a mentor with an expertise that is relevant to your project. The mentors’ role would be to give constructive feedback to improve your proposal. This feedback can be technical (e.g. pointing to some recent work that can help you improve your proposal) or non-technical (e.g. give you advice on how to improve your presentation).
After the Indaba week: The Indaba will support the 5 winning projects not only with seed funding, but also through our mentorship program in order to help turn your wonderful ideas into actual projects. The goal of this initial period is to turn the idea and the pitch into a well-structured research proposal, with initial studies and empirical investigations that can allow the winning teams to apply for other sources of funding:
- Financial support: Each of the winning teams will receive cloud compute credits of a minimum value of 500 USD.
- Mentorship: The Deep Learning Indaba has now a well established mentorship programme which supports short-term, transactional mentorship. This platform could be leveraged to connect you with experienced researchers to help you progress in your project definition and give you feedback on the research questions or obstacles you are facing at various points in your project. This mentorship can also have a non-technical nature such as helping you with structuring your research proposal or helping setting up a formal collaboration across the borders.
After one year from DLI2022, the 5 winning teams would be able to independently run their projects. They will also be invited to present their initial results and progress at future Indaba meetings.
Moreover, the Deep Learning Indaba’s support will continue beyond this initial phase. If a project is particularly successful (e.g. accepted to a top-tier conference or journal, or has a high societal impact), the Indaba will offer the team support to publicise their work at a wider scope.
Possible challenges and potential impact
Putting together an interesting and realisable research project is always a challenge. It is even more so when it brings together diverse parties. We are aware that these proposals could be faced with bureaucratic and political realities that makes their realisation
harder or less likely. Our hope is that we can work together to overcome these kinds of challenges, effectively building bridges not only for the competing teams to benefit from but all the machine learning communities in their countries. If successful, this experiment can go one step further towards building a more solid and connected machine learning community on our continent, in addition to the technical contributions it can yield.