How Time Travel Energizes Data Science Imagination

[#1]

What do multi-language machine learning deployment, text-based image enrichment, automated featurization, real-time predictive maintenance, reinforcement learning and biomedical image segmentation have in common? They are the topics that have been selected for this year’s Talk of the Geeks experience.

Why geekiness deserves to be celebrated

I dug into the word ‘geek’ a little last year. Historically the term geek comes from the German word ‘geck’ which literally translated means crazy.  Thankfully things change. Geeks are no longer the socially awkward, nerdy creatures residing in a basement afraid of being exposed to broad daylight only thinking in zeros and ones. Geeks are the new, high-impact drivers of our society.

They(we) are passionate, curious and with focus to make things work. If you are a geek, you know that it is in our nature to start working on an uber-idea and prove to anybody who wants to listen that it’s not merely an idea but something that actually works in practice. A powerful talent in this era of digital transformation.

What does time travel have to do with the advanced analytics landscape?

Turns out, a lot. We took inspiration from a childhood favorite, Professor Barabas. His many inventions often drive the story forward, even if the invention itself is not the story. We believe for many of our fellow geeks, advanced analytics may not be the star of the show in their business, but is key to driving transformation.

As is often the case, past inventions can inspire us to move further. Sir Isaac Newton famously proclaimed “If I have seen further, it is because I stood on shoulders of giants”. He was of course referencing the work of previous mathematicians that he was able to leverage. Similarly, past non-digital inventions are now giving us the opportunity to re-imagine possibilities.

Talk of the Geeks then is able to highlight how today’s innovations in data science are inspired by past technologies, and how this will impact our future. A virtual time travel experiment to educate and entertain.

Six must-see topics

We have curated topics that we believe data scientists would like to have a heads-up on today’s machine learning capabilities. Here are the ones we chose:

1. Polyglot or not: all languages are hot - Although there have been attempts to design one "universal" programming language in the past, we see today that this is a utopian dream. In reality, there is no need to choose one language, but to work with tools to manage different programming environments and integrate them with each other. In this presentation Kaat Tastenhoye will show you how you can build a model in Python and industrialize it in SAS.
2. Reimagining scans - Who would have thought that 100 years after Wilhelm Röntgen made the first X-ray image, CT scanners are using this technology to build 3D images of the complete human body? Can we use machine learning to automatically segment a CT scan into multiple classes like organs, bones or potentially dangerous lesions? Tune in and learn how with Joran Roor.
3. Let your pictures speak to you - How do we go from machines reporting “I see 30 people” or “I’m 60% sure I see a tiger” to describing what that tiger is doing? Learn how the power of Natural Language Processing can translate a picture into words as Jaimy van Dijk demonstrates Image Captioning.
4. Maintenance needs are more predictable thanks to the abundance of sensor data. But how to distinguish normal values from anomalies? By tracking mutual correlations using subspace tracking, anomalies and performance degradation can be found early such that costly unscheduled maintenance can be avoided. Rik de Ruiter will show us how.
5. Can you train your model like you train your dog? Reinforcement learning (RL) is an area of machine learning concerned with how software agents ought to take actions in an environment in order to maximize the notion of cumulative reward. From beating masterminds at playing games to real world applications, Joline Jammaers will show us how RL has it all. 
6. Automated Search to the Golden Feature Set – this is my talk, and I will introduce the Feature Machine, an automated featurization approach in SAS that will assess data quality issues, apply the appropriate feature transformation strategies to generate the golden feature set to improve data quality and consequently model performance.

See you on June 19th?

We aim to make this the best 60-minute experience in June for our community. See you online, on the 19th?