HTEC Deep Learning Meetups: Optimizing Algorithms

HTEC engineers Ugljesa Milic and Dusan Nastic held two meetups in Serbia on the subject of Deep Learning, specifically on improving algorithms for mobile processors.

The 7th HTEC TechTalk in Nis was held briefly after the gathering in Novi Sad, where they discussed the same topic. On both occasions, Ugljesa and Dusan tried to explain why there is a great need to run deep learning apps on mobile devices, but above all else what the significance of this unique project being developed in Serbia is.

Other topics included the demonstration of Halide programming models as a way of coding and optimization of Deep Learning and Image Processing algorithms; the potential of machine learning and impact it can have on the further evolution of mobile processors, which will potentially ease their programming and speed up the algorithm execution with lower power consumption.

Deep Learning Meetup Halide Programming

“If you ask me, both meetups went great. We spoke about specific things related to Machine Learning on mobile processors, and although we were afraid the topic would not be broad enough for the listeners, it turned out that the audience received it well and they fully understood the story we wanted to share”, Ugljesa told us after the meetups. He added, “The audience in Novi Sad had some amazing questions after the presentation, and in Nis, we were thrilled to see how diverse the audience was, from freshmen to experts. We were pleasantly surprised by the number of people who dealt with similar problems to ours. All in all, this was an amazing experience, and we can’t wait to go back to Novi Sad and Nis and talk about some new interesting things.”

Ugljesa Milic is a Machine Learning engineer in HTEC and has a Ph.D. from Barcelona Supercomputing Center, where he worked on micro-architectural processor optimization used in High-Performance Computing. He also has work experience with companies such as Cray and NVIDIA.

Dusan Nastic has an MA in electronics from the Faculty of Electrical Engineering Belgrade but has on his own started research in the field of Machine Learning as he has always been interested in image processing and ML.

If you would like to hear more from Dusan and Ugljesa on the subject of Deep Learning, you can listen to our Protok podcast (in Serbian) and learn about the history of Deep Learning, the difference between DL, ML and AI, self-driving cars, marketing potential, voice recognition, medical products, etc.

If you liked this article

Read more related posts:

|Tech Blog

Building Intelligent Routing Models

The significant expansion of crowdsourcing applications in the last decade, along with the vast availability of mobile phones and similar personal sensory devices, has inevitably generated a considerable amount of location data. The ability to transform locations into business value has triggered the interest of our data-driven economy in spatial data analysis. Geo data science has tapped its way into the emerging field of data-related solutions, expanding the reasoning behind such solutions to an additional dimension – a spatial dimension.

Read More
|Tech Blog

Machine Learning in Medical Data Analytics

Over the past few years, deep learning practitioners in both the industry and academia have reported state-of-the-art performances across many tasks in various fields, and a significant portion of them even surpass the human-level performance. Deep learning found its application in many areas of healthcare. It can be applied to medical image diagnosis, drug discovery, robotic surgery, and others.

Read More
|Tech Blog

Machine Learning – Halide Programming Insights

In the last decade, Moore’s Law continued to provide more and more transistors per area unit. In the beginning, computer architects used those transistors to increase single-thread performance, designing power-hungry central processing units (CPUs).

Read More