Experiences at Indabađť•Ź Kenya: Data Driven Patient Diagnosis with Dr. Elsa Poster Presentation
Over the weekend of 7th April 2018, my team and I were invited to the IndabaX Kenya event, organized by the wonderful Women In Machine Learning (WIML) group, to present a poster and share on our AI powered Health and Telemedicine project and the experience was truly nothing short of inspiring.
In this post I will talk a little about what went on during the event, what I learnt, and what I was able to share with the great vibrant community of Nairobi!
The Event
Indaba — Strengthening African ​Machine Learning — and the Indaba events are a great effort by African innovators and thought leaders to make more Africans an important part of the current Artificial Intelligence era.
A Deep Learning Indabađť•Ź is a locally-organised, one-day Indaba that helps spread knowledge and builds capacity in machine learning.
We kicked off with the keynote: The ML Roadmap for Kenya: Where We Are And Where We Are Heading by Nikhil Ravichandar, where he talked about the state of Machine Learning in Kenya and Africa as well as what the next steps should be in the journey towards more pervasive machine learning solutions.
We then went on to the pannel session, the practical session, lunch and the poster presentations.
The Panel Session: Artificial Intelligence, Drivers and Forces at Play
This was easily one of the best parts of the day, to see a panel of very achieved industry leaders discuss the whats and the hows of Artificial Intelligence, covering everything from policies and governments to the ethical concerns with AI and data. It was a great learning experience for all who attended.
Lead by Professor Bitange Ndemo, this session was filled with humor and thought provoking ideas on humanity and where we are headed. It was also very nice to see how the different panelists had different ideas on how to address challenges and even different ideas on what constitutes a problem. This was one of those “You had to be there to know what I mean” things.
The Practical Session, and What I have learnt:
Led by the knowledgable Brian Muhia with the support of Amina Islam and Kathleen Siminyu, the practical session focused on the more technical and “practical” aspects of machine learning. We saw the tedious process of data cleaning and preprocessing, the various types of regularization like Dropout, and how easy it is to use PyTorch to set up a deep neural network to Learn Entity Embeddings of Categorical Variables.
I learnt a few things on using the Fast ai modules and an interesting comment on “Drop connect” as an alternative to Dropout. All in all, it was
great to see the interest from the community to learn more on this.
The Poster Presentations
After lunch and a bit of networking we were ready for the next part, poster presentations. For this part, the Women in Machine Learning (WIML) group had invited people from East Africa, Tanzania included (us!), to come and share what they are working on and what their findings are. The event organizers turned the poster session into a friendly competition for the grand prize of 5,000 Kenyan Shillings!! The stakes were high, and the tensions were even higher!
We saw some great presentations on using computer vision to detect driver drowsiness, using decision trees and KNN algorithms to predict a loan seekers probability to default on their loans without using any bank history and understanding opinions in text. We could not get enough of this!
I presented on Dr. Elsa, our AI backed health and telemedicine service that is offered for free in Tanzania. We empower doctors to make smarter and more informed decisions and provide patients with a safe and secure way to consult a doctor free of charge.
We talked about how we are using deep neural networks and gradient boosted trees to provide a differential diagnosis of patients on infectious diseases in Tanzania, as well as our work with the National Cancer Research Institute to develop tools that will help with the early detection of cervical cancer. The crowd was super supportive and the judges very encouraging in their questions and suggestions.
Our presentation ended up winning the competition and the grand prize of 5000 Kenyan Shillings!! Whaaaaat???? More importantly, from one of the judges responses, judge Loki from Safaricom, actually helped us rethink an aspect of our ensemble model and are expecting to see improvements in accuracy after the changes.
Special Thanks and Shoutouts!
On behalf of everyone who attended the event, I would like to thank the Busara Center for Behavioral Economics for the wonderful venue and the panelists for sharing their knowledge and experiences in the industry.
I can’t say it enough so I’ll say it again, the Women In Machine Learning group, WIML, did an amazing job at organizing and running this event, I will be honored to be invited to attend the next one! Also shoutout to the lively Muthoni Wanyoike, and the gracious Kathleen for being such wonderful hosts!
