Comtrade Digital Services Platinum sponsor at Wonderland AI 2019
POSTED ON 28.11.2019
Have you heard about Wonderland AI? It is a conference with the aim to promote artificial intelligence.
This November, Comtrade Digital Services joined Wonderland AI Summit, as Platinum Sponsor.
The one-day Summit gathered a variety of speakers, including the ones from IBM, Mercedes Benz – Daimler, Hewlett Packard Labs, Facebook and many more. More than 600 people innovators, managers, engineers were present at the Madlenianum Theatre on 8th November.
During the exhibition period, we had many interesting talks at CDS booth. We presented the concept of our AI Collective and exposed services we provide in AI domain to the broader audience coming from business and academia.
Beside exhibition part, our Data Science/AI researcher, Aleksa Radosavčević, has discussed some of the theoretical concepts of ML and its automation potential, while the focus was on two practical examples where AutoML implementation led to more robust and optimized solutions.
Aleksa had amazing presentation which was attended by more then 100 people.
If you haven’t had chance to hear it live, you can take a look at very witty description for you:
Each data scientist, at each step of the ML process, is faced with choices and alternatives. alternatives space is almost limitless. Imagine standing in center of the Earth, and you need to point your finger to the dot on the surface, size of a tennis ball, representing an optimal ML configuration. Now think how many of these tennis balls would be needed to cover Earth’s surface.
Roughly speaking, if overall ML process is divided into three parts: data-flow optimization, features engineering & selection and algorithm hyperparameters optimization, we could speak about “perfect storm” that data scientists are almost forced to pass to arrive at the solution. AutoML is the best possible boat to jump in when passing through this storm.
Recent developments in AutoML field not only facilitated and speeded up the process, but also allow less technical people – “citizen data scientists” to get more deeply involved in data science and machine learning tasks. In this manner, the focus is shifting to employing in-domain knowledge, rather than experimenting with algorithms. But on the other hand, higher levels of abstraction could lead to ignorance with respect to what is going on “under the hood” and how the solutions are created in the first place.
Therefore, finding the right balance between tailor-made AutoML solutions and publically available toolkits is essence for successful implementation of ML projects.