Tue, May 7, 2019
In January, I got the great news that I had been invited to give a talk at the sold out Applied Machine Learning Conference (AMLC), which would take place at the Tom Tom Summit & Festival in Charlottesville (Virginia) in April. I had heard great things about the 2018 edition and was eager to know Charlottesville, so I immediately accepted the invitation. 🤗
The Tom Tom Summit & Festival is an annual event with the following mission:
Celebrating entrepreneurship, culture, and innovation in Charlottesville and inspiring small city excellence everywhere.
Along with the AMLC, Tom Tom Summit & Festival consists of 5 other conferences: Youth Innovation, Civic Innovation, Renewable Energy, Creative Ecosystems, and Entrepreneurial Ecosystems.
Charlottesville was a great location for the summit. The city of only 48,000 inhabitants is becoming a small hub for artificial intelligence and machine learning research and practice, boasting the presence of organizations such as Astraea, Metis Machine, and S&P Global, as well as the Data Science Institute at the University of Virginia and National Ground Intelligence Center (NGIC).
Currently, I'm working as a research engineering at Tryolabs, focusing on computer vision. So, I thought it would be fun to talk about Convolutional Neural Networks (ConvNets or CNNs), given that they are the backbone of every algorithm I've been using lately.
These were the main aspects of CNNs that I covered in the presentation:
Check out the slides of my presentation here:
Automated machine learning is one of the most interesting fields in artificial intelligence, promising to facilitate the job of Data Scientists and Machine Learning Engineers, or even take over the creation of new learning architectures. In this talk, Adam Blum highlighted the different machine learning tasks that can be automated, including the selection of algorithms, search of hyperparameters, cross-validation, data preprocessing, and feature creation.
In this talk, Andrew Bollinger explained how he and his team used machine learning to help manual annotators create a map of the high voltage electrical grid in Pakistan. They used satellite data to train CNNs, which they then used to predict the location of the high voltage lines. The creator of the map can then use this data as a guide when mapping the electrical lines. This greatly assisted in reducing the time it takes to create a country wide map, which is a huge task.
When drawing up charts I usually just focus on making them intelligible and don't ever really think about their aesthetics. This talk really opened my eyes on how we can use visual metaphors to not only make the charts more visually pleasing, but also make them easier to understand and memorable by making the viewer connect concepts they already know with the new idea the chart is trying to explain.
Are you interested in attending a machine learning conference but you missed the Tom Tom Festival? Check out our list with machine learning and deep learning conferences coming up around the globe. It includes notable speakers and discount codes.