Autoencoders are neural networks that are trained to replicate input using less information. In this presentation we explore the use of autoencoding to learn about latent factors that drive security returns. We extend this approach to create a dynamic autoencoding model. We apply these approaches to both simulated and actual financial return data sets and present the results.
Scott has 20 years of financial markets experience with investment banks and hedge funds. He is currently working with Noviscient (www.noviscient.com), a technology-led, research and proprietary trading firm based in Singapore. Noviscient applies statistical and machine learning technologies to investment management challenges. Prior to this he was Chief Risk Officer and Portfolio Manager at Vulpes Investment Management. Scott also spent eight years with Deutsche Bank in Singapore heading a quantitative team covering modeling and valuation risk for the Bank’s Asian trading businesses. And before that he worked in venture capital with Macquarie Bank in Sydney. Scott has a BEng and an MBA from the University of Melbourne, a Master of Quantitative Finance from the University of Technology, Sydney and he is completing a PhD in Finance with EDHEC. Scott has been coding in Python for around eight years and working with machine learning techniques for five years.