Strategies in low-data regimes

We begin the new year with a new meetup. This time David and Marcus from Tenfifty are hosted by Valon at Telia.

Bayesian belief networks for human dialogue
In many fields, we do not have enough data to create complete models from scratch. With belief networks, we can code expert knowledge, including uncertainty, and gradually improve knowledge by adding data. The network can further help decide what sort of information would be most valuable to ask for next in a dialogue system.
David Fendrich

Object localization in the low data regime
Image processing is one of the fields in machine learning that has seen the greatest progress in recent years, to a large extent explained by the ever-growing datasets available. But how well does the state-of-the-art transfer to cases where data is limited? How much work is required to bring such a project from conception to production, and what techniques can help on the way?
Marcus Olivecrona

Tenfifty delivers custom-built subscription services using AI, data science and crawling. Whether you have text in any language, voice, image, or ordinary numerical data, we have deployed solutions into production for two decades.

Agenda
17:30: Food and drinks
18:00: Presentation
19:00: Mingle
20:00: Closing

Note: We start 30 minutes earlier than usual due to venue restrictions!