Discente: Alan Rodrigues Fontoura
Título: A Deep Reinforcement Learning Approach to Asset-Liability Management
Orientadores: Eduardo Bezerra da Silva (orientador) e Diego Barreto Haddad (CEFET/RJ) (co-orientador).
Banca: Eduardo Bezerra da Silva (presidente), Diego Barreto Haddad (CEFET/RJ), Laura Silva de Assis (CEFET/RJ) e Aline Marins Paes Carvalho (UFF)
Dia/hora: 22 de julho de 2020, às 14h.
Sala remota: meet.google.com/jox-
Asset-Liability Management (ALM) is a technique to optimize investment portfolios, considering a future flow of liabilities. Its stochastic nature and multi-period decision structure favors its modeling as a Markov Decision Process (MDP). Reinforcement Learning is a state-of-the-art group of algorithms for MDP solving, and with its recent performance boost provided by deep neural networks, problems with long time horizons can be handled in just a few hours. In this work, an ALM problem is addressed with an algorithm known as Deep Deterministic Policy Gradient. Opposed to most of the other literature approaches, this model does not use scenario discretization, which is a significant contribution to ALM study. Experimental results show that the Reinforcement Learning framework is well fitted to solve this kind of problem, and has the additional benefit of using continuous state spaces.