Beyond Human Labeling and Supervised Learning w/ Google

[VIRTUAL REPLAY] Bringing data enthusiasts together to foster the exchange of ideas and the intellectual growth of the data community: Dataiku series of meetups showcase the work of talented data professionals across industries, for you to get insider tips and tricks to turn data into actionable insights (read more here (https://blog.dataiku.com/announcing-dataconnect-meetups-nyc). Typically most companies have been using supervised learning with human labeled dataset. However, thirst for more labels is never satisfied, especially for deep learning models. In this talk, Karthik will talk about technologies to overcome lack of abundance of human labeled dataset with specific examples in computer vision and NLP domains. The talk will explain how synthetic dataset can be used to train a model that can be generalized to operate on real world data. Then, Karthik will cover self-supervised learning that is becoming an emerging trend in large scale deep learning tasks. The talk will focus on a couple of use cases of self-supervised learning techniques that are general enough to be applicable in the majority of computer vision tasks.