Attention in Multi-Task Learning (MATTE)
- Gjorgji Strezoski, Ph.D., University of Amsterdam, Amsterdam, The Netherlands

Abstract: Attention has been a recurring topic in Computer Vision regardless of architecture or methodology choices. It allows for implicit control over the model's resources and accents the insight-heavy data regions. In this talk, we are going to navigate the basics of attention and discuss a general method for learning task and scale-based attention representations in Multi-Task Learning (MTL) for vision dubbed MATTE. The discussed approach relies on learning and maintaining cross-task and cross-scale representations of visual information, whose interaction contributes to a symmetrical improvement across the entire task pool. MATTE adds an attention-modulated feature as residual information to the processing of each scale stage within the model, including the final layer of task outputs. As the primary area of application for MATTE lies in dense prediction tasks such as image matting, we are going to observe the effects of this type of attention approach over real live data.

How to Build an AI Startup from Scratch
- Svetlana Kordumova Trajanova, Ph.D., Founder and CEO at Pixyle.ai, Part-time Assistant Professor at UGD, Faculty of Informatics

Abstract:Building a startup from scratch takes more than just an idea. The most important first step is taking an action upon that idea, and then taking lots of other actions upon the first action. Pushing, gathering feedback, taking every criticism as advice, pivoting, changing, and growing. Building a startup changes you as a person. Being open-minded, and open to change is the first crucial characteristic for any entrepreneur. Join me in sharing my journey, and I hope to inspire a number of you to take this direction of building a startup or joining to work for a startup. It’s a hell of an exciting ride.

Financial Robo Advisors, Machine Learning, Data Lakes, Fairies and Unicorns – Banking Technology Triumphs and Pitfalls
- Bojan Najdenov, M.Sc., Senior Advisor at 44 EAST, Copenhagen, Denmark

Abstract:The banks today globally exist in a highly complex environment, not only from an organizational perspective, business processes, and service offerings point of view but also from the nature of the challenges that the sector is exposed to is highly complicated as well (e.g. overcoming legacy systems, utilizing new technologies to maximize business value creation, constant treats from FinTech start-ups and digital front-runners, simplifying and scaling of a business process). Furthermore, they are affected by constant changes in the macroeconomic climate, threats of inflation and recession, volatile markets, high cost-over-income ratios, strong regulatory pressure, and demanding digitally-native customers. How can the advances in information technology (machine learning, big data management, distributed computing, and management of information systems) enable banks to keep up with their competition and lead the way to the future?