GTC Silicon Valley-2019 ID:S9743:Factor Investing Using Deep Learning
Rafael NicolasFerminCota(RN Financial Corporation),Steven Thornton(RN Financial Corporation)
We'll talk about how we use deep learning and GPU-Accelerated portfolio construction techniques to generate a long-short portfolio. We start with a database containing more than 4,000 daily factors on more than 6,000 publicly traded U.S. equities over nearly 30 years. We'll explain how we apply deep learning to process this data and identify relationships that forecast relative equity performance at multiple time horizons. Our neural network identifies long-short portfolios that we combine into our final portfolio by using a CUDA implementation of a risk-parity algorithm.