On October 14, the HTP business incubator hosted AI Day Minsk, a conference to support the global initiative started in London to unite the global data communities through series of events and meetups – Big Data Week (October 9-15, 2017).
The event featured speakers from InData Labs, Wargaming Minsk Team, and Flo: Smart Period Tracker. Leaders of the growing Belarus' AI community spoke about introduction of AI technologies into existing projects, use of deep learning technologies for application problems, and the latest machine learning methods. The event was attended by employees of the leading Belarusian IT companies, scientists from the United Institute of Informatics Problems of the National Academy of Sciences of Belarus, students and teachers of the Belarusian State University of Informatics and Radioelectronics, and representatives of the AI communities in Russia and Ukraine.
The speakers prepared six lectures of different levels of complexity aiming to cover various mathematical, technical, and business aspects of artificial intelligence and big data development.
Dzianis Pirshtuk, Chief Data Scientist at InData Labs, addressed machine learning and artificial intelligence as a business tool and a foundation for new products. He spoke about the machine learning development history and trends as well as the new opportunities it provides to solve urgent social and economic challenges.
Roman Merkulov, Data Scientist at InData Labs, provided an overview of approaches to time series analysis, described the application conditions and restrictions of different methods, and compared tools useful in time series forecasting.
Andrew Kovzel, CTO at Ideanomix Digital, and Denis Dus, Data Scientist at InData Labs, shared their experience in using neural networks to improve cycle predictions (Flo success story).
In his lecture on multi-label classification in natural language processing (NLP), Alexei Tishurov, Data Scientist at InData Labs, focused on sources and examples of NLP tasks, peculiarities of preliminary analysis, metrics choice, and advantages and disadvantages of different approaches to solving NLP tasks.
Levon Avakyan, Server Reliability Team Lead at Wargaming, shared World of Tanks experience with Apache Kafka as a message broker. He described the Kafka cluster, explained what tasks can be solved using Apache Kafka, and revealed what challenges the Server Reliability team faced and how they were overcome.
The last lecture of AI Day Minsk was delivered by Denus Dus and dealt with probabilistic modeling in deep learning. The speaker outlined the advantages of using probabilistic modeling when writing machine learning algorithms, explained how probabilistic modeling and regularization methods in neural networks are related, and considered ideas of a Bayesian world in the context of deep learning.
During the breaks, participants shared their experiences, discussed the lectures and topical issues in the fields of artificial intelligence and big data.
The lectures are available on YouTube.
You may see the album on Facebook.