Machine Learning Prague! I was there and I'll tell you how it was.

Machine Learning Prague 2019. Source: Author. 

Last weekend I had the opportunity to attend one of the biggest events of Machine Learning in the world: Machine Learning Prague. My participation took place not only as an attendee, but also as part of the Microsoft Team, supporting and promoting Microsoft technologies. Now, I have to say that I was in a privileged position, considering that the tickets cost 200 Euros each. Therefore, I endeavoured in getting the most of it and I'd like to share with you my remarks on the conference.

The Venue and Organization

The Conference was held in the beautiful city of Prague, in the building Rudolfinum, next to the river Vltava. Rudolfinum is an architectural jewel and is now the seat of the Czech Philharmonic. The organizers chose a really fancy place to hold the conference.

Rudolfinum, Prague. Source: Author. 

The Program

The conference's program was comprised of three days full of activities, including parties at night! The first day was dedicated exclusively to workshops, which unfortunately I couldn't attend. Nonetheless, the official conference's agenda started on Saturday and we had different sessions starting from 10:00 until 18:00.

From the first day, I remember particularly the last two speeches. First, the speech from SpaceKnow who are using ML and GANs, in order to process Radar Satellite Imagery. How? - you might ask. Well, they are using ML to "remove clouds" from satellite imagery, so that they know what's going on on earth at any given time. Second, the speech from IBM. They are using ML to identify vehicles in satellite images. However, the problem is more complex than it seems. What if vehicles are partly covered? What if they are under a shadow? Or under a tree? IBM uses ML in order to process the images, using different algorithms. Each algorithm produces an image which produces special patterns for vehicles, helping with its identification.

Jan ZikeŇ° from SpaceKnow. Source: Author

The second day started and ended one hour earlier as of the previous day. We had several speeches being the most interesting for me the ones from:
  • Spotify: Marc Romeyn from Spotify came and talked about how they are using ML and Tensorflow to deliver the awesome service they provide. Specifically, they are using ML in vectors to capture the semantics. Consider the following example: 
    king - man + woman = queen
    They used this example to explain how Spotify makes a choice of the songs for our "Discover Weekly". Pretty clever!
  • or the so-called "Czech Google". It is one of the most popular search engines and mail servers in the Czech Republic. They explained to us how they use ML in order to process natural language, create semantically correct sentences, and look for results in their databases. The algorithm gets a little bit more complicated when you have to deal with Czech language and its 7 declensions. The search engine of captures the user's input, processes it, validates its correctness, provides suggestions, and forecasts the next words/phrases, just as Google does. 
Marc Romeyn from Spotify. Source: Author. 

Main Takeaways

Now, what are my main takeaways from this conference?
  1. Python and Tensorflow are on the lead.

    When it comes to Machine Learning, it seems that Python and Tensorflow are leading the path as the most popular language and framework for ML development, respectively. Python has come a long way, from being just another regular programming language to becoming the elite for Big Data, Machine Learning and AI. And all thanks to the support of thousands of developers who created the specialized libraries that Python nowadays has. On the other hand, Tensorflow, the open source machine learning framework created by Google has acquired a lot of popularity among developers and among the largest companies in the world. Big players like Spotify, Coca-Cola, Airbnb, intel, SAP, etc. are implementing this framework. Furthermore, most of the companies that participated in ML Prague talked about using this framework on their own projects. Well, it seems like it is a good time to start learning a new technology.
  2.  It's not only for developers.

    During the conference, I had the chance to meet not only developers; but also mathematicians who were implementing and developing new algorithms, sales people who were trying to find out new use cases for the available technologies, analysts considering the impact of AI in the world, students who wanted to learn about Machine Learning and technology enthusiasts. Certainly, ML is a broad topic that will affect the big majority of economic fields. Therefore, there is a good chance that you might encounter it throughout your career, even if you're not a developer.

What about Microsoft?

As a Microsoft Student Partner, I am aware of all the efforts that Microsoft is putting on developing the Microsoft Cognitive Toolkit. In a future article, I'll write about how you can start using Microsoft Azure Cognitive Services in a playful and fun way.

Me at the Microsoft Booth at ML Prague 2019. Source: Author.

Now, for closing I'd like to say that what's important is to start tinkering! Download the tools, do small projects, get yourself involved. Do whatever you want, but tinker, tinker, TINKER!