The Latest in AI @ bol.com #2

17.30 – 18.30 Welcome, food and drinks
18.30 – 19:00 Mariya Hendriksen - Multimodal AI for Retail
19:00 – 19:30 Samaneh Heidari - Social Norms
19:30 – 19:45 Break
19:45 – 20:15 Valentin Calomme - Automated Machine Learning: a realistic overview
20:15 – 20:45 Willem Vlot - GeoAI: accelerating location intelligence using machine learning
20:45 – 21:30 Drinks

Parking is possible at our office.

Mariya Hendriksen - PhD Researcher, AIRLab / bol.com
Multimodal AI for Retail.
Samaneh Heidari - PhD Researcher, Utrecht University
Social Norms
Social phenomena are part of our thinking. Therefore, it is mandatory to consider social aspects to study decision making and system behavior. Among different social aspects, we are interested in studying social norms, as norms play an important role in guiding all human societies. Social norms are more important to study and consider in the absence of a central monitor/control. Social norms are important as societal agreements of acceptable behavior. They can be seen as flexible, but stable constraints on individual behavior. However, social norms themselves are not completely static. Norms emerge from dynamic
environments and changing agent populations. They adapt and in the end also get abrogated.
In the presentation, I will talk about social norms, multi-agent simulations, and the applications of normative multi-agent simulation.

Valentin Calomme - AI Engineer, Mediaan
Automated Machine Learning: a realistic overview
In the past few years, Automated Machine Learning has become a big topic of conversation. AutoML brought the promise of allowing virtually anyone to build competitive machine learning models with little to no programming or machine learning expertise. Who would not want that? Some even predicted that it would be the end of data scientist jobs altogether. After all, no one is safe from automation. Needless to say that these predictions didn't prove to be prophetic. AutoML has made great strides and has challenged the status quo. However, it does have many limitations, some that are structural and unlikely to go away, and some that may be removed in the upcoming years. In summary, Automated Machine Learning has a place in the machine learning lifecycle, but it is merely not a plug-and-play replacement that shatters how machine learning is being done, at least not yet. This talk discusses what AutoML is and what is currently available for developers; for instance, SMAC, BOHB, Auto-Keras, or even TransmogrifAI. It also covers when one should or should not use AutoML today, and what might change tomorrow.

Willem Vlot - Analytics Engineer Esri
GeoAI: accelerating location intelligence using machine learning
One area where machine learning thrives is within location intelligence, the capacity to organize and understand complex data through the use of geographic relationships. Massive amounts of data are linked to a physical location and/or a moment in time. Using data-driven algorithms and techniques that automate prediction, classification and clustering of geospatial data, organizations uncover hidden insights-the kind of information intelligence that creates a crucial competitive advantage.
This talk delves deeper into the leading edge of geospatial AI capabilities and their impact on decision-making.