Portfolio optimization focuses on risk and return prediction, yet implementation costs critically matter. Predicting trading costs is challenging because costs depend on trade size and trader identity, thus impeding a generic solution. We focus on a component of trading costs that applies...

This thesis presents the development and implementation of a novel Deep Distributional Reinforcement Learning (DDRL) approach in the field of quantitative finance: the Distributional Soft Actor-Critic (DSAC) with an LSTM embedding. The model is built to further stabilize the performance of the...

Covariance matrix forecasts for portfolio optimization have to balance sensitivity to new data points with stability in order to avoid excessive rebalancing. To achieve this, a new robust orthogonal GARCH model for a multivariate set of non-Gaussian asset returns is proposed. The conditional...

Recently, a large body of academic literature has focused on the area of stable distributions and their application potential for improving our understanding of the risk of hedge funds. At the same time, research has sprung up on standard Bayesian methods applied to hedge fund evaluation. Little...

We argue that the practise of valuing the portfolio is important for the calculation of the VaR. In particular, the seller (buyer) of an asset does not face horizontal demand (supply) curves. We propose a partially new approach for incorporating this fact in the VaR and in an empirical...

This paper presents mathematical formulation for the probability distribution function of asset value difference using canonical ensemble framework. For asset value significantly smaller than the total market value, the distribution is given by exponential function, which depends on market-eta...

Under an assumption of normality, we explore a non-orthogonal Bayesian technique in which redundant information can in principle be filtered out of the posterior distribution by the explicit coupling of the prior and likelihood functions. The Black-Litterman forecasting model widely used by...

In this work, we propose an ARMA(1,1)-GARCH(1,1) model with standard classical tempered stable (CTS) innovations for historical daily returns of 29 selected stocks. The non-Gaussian nature of the innovations captures the fat-tail property observed in data. The dependency between different assets...

This paper constructs Tail Risk-Managed (TRM) portfolios in real time, where the scaling of exposures to factors is determined by forecasts of probabilities of VaR violations. Using a set of US Fama-French factors and a set of International equity portfolios, we show that TRM portfolios achieve...

Predictive power has always been the main research focus of learning algorithms with the goal of minimizing the test error for supervised classification and regression problems. While the general approach for these algorithms is to consider all possible attributes in a dataset to best predict...