全球-量化策略-深度学习机器人交易员1.0：应用递归神经网络预测最优投资组合配置.pdfVIP

Global Quantitative Derivatives Strategy 24 September 2019 Deep Learning Robot Trader 1.0 Applying Recurrent Neural Networks to Predict Optimal Portfolio Allocations Hi Siri, how much should I trade? Global Quantitative and We train deep neural networks to directly predict optimal portfolio Derivatives Strategy allocations, with an objective to maximize strategy performance such as Ada Lau AC Sharpe ratios. Unlike traditional investment processes which firstly predict (852) 2800-7618 returns (alpha generation) and then control risk (portfolio construction), it ada.lau@ could be more natural and efficient to directly predict optimal portfolio J.P. Morgan Securities (Asia Pacific) Limited/ allocations that deliver good risk-adjusted returns J.P. Morgan Broking (Hong Kong) Limite Marko Kolanovic, PhD Maximizing the Sharpe ratio of an Equity/Bond portfolio (1-212) 622-3677 We look at an architecture of recurrent neural network called the Long Short- marko.kolanovic@ J.P. Morgan Securities LLC Term Memory (LSTM) model, which is