Abstract: In this talk we explore how in context encoding algorithms starting with the breakthroughs over the past 18 months impact real world use case applications. We will do a deep dive into transformer based architecture with technical focus on the registerables implementation paradigm and compare performance with prior art for automated “Fake News” evaluation using contemporary deep learning article encoding. We explore how these techniques provide unique interpretability for the FakeNews use case and close with a discussion of extensions of these techniques to time series forecasting and telemetry monitoring.
Bio: Coming soon!
Data Scientist | Symantec
advanced-w19 | deep-learning-w19 | intermediate-w19 | talks-w19