SEMINAR CANCELLED Deep learning made easy with TerseNet
LocationErasmus University, Mandeville Building, Room T18-25b
Date and time
March 17, 2020
13:00 - 14:00
Neural networks show promising predictive performance in economic and financial applications, but the final model is hard to interpret, model selection is an arduous and time-consuming task, and the model behavior becomes unpredictable during rare events. We propose the Tesselated regression simplex encoding Network (TerseNet): A framework with straightforward model selection, predictable behavior for rare events, interpretable models and an overall lead time that is orders of magnitudes shorter than ordinary neural networks.