Nikola Rahman, HTEC’s machine learning engineer, showcased HUMEDS, a telemedical solution intended for ECG signal acquisition and detection of heart arrhythmias. Nikola introduced the audience to the issue of atrial fibrillation, a common type of arrhythmia which affects about 33 million people around the world, and how his team employed convolutional neural networks to detect it. The presentation went over the employed dataset, the model used by the network (a 34-layer ResNet), as well as the preprocessing and network training methods. There was also talk about the next steps in the development, which include further research and development, and porting of the network to mobile devices.
The second presentation featured Stefan Jovanovic, a data scientist from Nordeus, who introduced the system of chatbots used by the tech support team of their hit video game, Top Eleven. Stefan talked about automatization of the common FAQ process in what is essentially a linguistic problem – fetching an answer to an inquiry by the user through a pre-trained neural network which recognizes phrases used in the inquiry.
Both presentations were topped off with the questions from the audience, which was primarily interested in neural network models used in both cases.
This Data Science meetup was the 23rd of its kind, which once again emphasized the vast interest in the area which has been shaping the IT landscape for the past several years. Along with our own Tech Talks, HTEC’s engineers will return to the spotlight of Data Science Serbia’s meetups in the spring next year, so be sure to subscribe to some of our social networks channels, if you’re interested.
Get in touch with the data science community in Serbia by following their meetup page.