Format:
- 2 presentations (each approx. 30-40 minutes including discussion)
- Of course time for networking + food + drinks before, in between and especially after the presentations
- Talks are held in English
The line-up:
Daniel Peterson – Understanding and Expanding Word2Vec
Daniel Peterson is a data scientist at TrustYou, and a PhD candidate at the University of Colorado. His work focuses on deriving meaningful insights from text. His first degree is a BA in Mathematics, which established the foundation in problem-solving that led him to machine learning. He's from Wyoming, the smallest-population state in the USA, and in his spare time he homebrews beer, reads, walks outside, and in the summers works in his vegetable garden.
Word2Vec has proven useful at many tasks in NLP, because the vectors learned capture many meaningful relationships. This has spawned a large number of variations suited to accomplishing distinct tasks – not unlike LDA did ten years earlier. In this talk, I'll walk through briefly how Word2Vec functions, its assumptions and shortfalls, and some simple, successful variants.
TBA – TBA
Here you can find more information about the Munich Datageeks and the event.
