Skip to main content Skip to navigation

Yijie Zhou

I am currently a 1st year Phd student in Mathsys II (funded by CSC).

I'm interested in complex networks, deep learning, reinforcement learning and related topics and applications in finance. My PhD project is supervised by Prof. Robert Mackay and Dr. Sansom Bazil.


  • 2020-, University of Warwick, Covnetry UK

PhD candidate in Mathsys.

  • 2019.09-2020.09, University of Warwick, Coventry, UK

Mathsys, Centre for Doctoral Training. MSc level (Distinction)

  • 2018.09-2019.09, Univerisity College London, London, UK

MSc Financial Risk Management, Department of Computer Science (Distinction)

Dissertation topic: Evolution strategy and its combination with deep reinforcement learning in portfolio management.

  • 2014.09 - 2018.07, Xi'an Jiao Tong Liverpool University (XJTLU), Suzhou, China

BSc Applied Math, First Class with honor

Final year project: Numerical methods with its application in finance.

  • 2011-2014, Suzhou High School of Jiangsu Province


MSc Individual project (2020)- Trophic analysis on cross border exposure network.
  • Research on Financial systemic risk has attracted interest from different disciplines since the financial crisis 2008-2009. As a complex system, the endogenous network structure of the financial system plays key roles in explaining the micro-macro feedback phenomena in systemic risk. Loop structure within the financial networks can as a risk amplifier. It has been identified as a pathway towards financial instability. Trophic incoherence, a term based on trophic level originally from ecology, quantifying the extent of the cyclicity of networks is related to the stability of a large complex ecosystem. A recent improvement in the definition of trophic level overcomes the methodology challenge and makes it possible to bring this network analysis tool to the field of financial risk. In this work, trophic level related analysis is applied to the cross border exposure network. This empirical network has been more trophic incoherent after the financial crisis. Besides, selected contagion models are used to quantify the loss given a shock to networks with different trophic structure. We show that a highly incoherent network is more vulnerable in high-risk scenarios. This work is the first of its kind to apply trophic analysis to financial networks and look into the relationship between trophic incoherence and financial instability.
MSc Research Group Project (2020) - Stochastic parareal: an application of probabilistic methods to time-parallelisation
  • An investigation into whether probabilistic methods could be incorporated into the parareal algorithm in order to extract further parallel speed-up.
  • Supervised by Dr. Massimiliano Tamborrino, Dr. Debasmita Samaddar and Dr. Lynton Appel.
  • Group member: Kamran, Jack, Jimmy, Haoran.
MSc Financial risk management individual project (UCL) (2019)-Evolution strategy and its combination with deep reinforcement learning in portfolio management.
  • Advanced machine learning techniques such as deep reinforcement learning techniques are widely applied in many fields. Therefore, this motivates the author to apply deep reinforcement learning (DRL) in portfolio management. However, existing DRL algorithms all use gradient-based method to optimize parameters which is time-consuming and suffer from the problem of the non-existence of derivatives. Evolution strategy inspired by natural evolution is proved to be a competitive alternative approach in many deep reinforcement learning tasks.
  • This master dissertation combines deep reinforcement learning and the evolution strategy in the field of financial portfolio management. This work presents a novel approach to allocate weights of different assets in a portfolio given a time period of trading. The core of the project is to find an optimal policy network for reinforcement learning agent. While the gradient-based method is used to find the optimal solution in the deep reinforcement learning algorithm for portfolio management, we attempt to use the evolution strategy to do exploration in parameters space. Combination with evolution strategy provides a possible alternative approach in deep reinforcement learning for portfolio management.



(2020-2021 Academic year) TA for MA3K1 Mathematics for ML and MA124 Maths by computer.


yijie dot zhou at warwick dot ac dot uk



(Monday/Tuesday due to Covid restriction)