Curriculum Grant sponsor Supervisor
s Co-Supervisor
s Contact | ||||||
Project description
Memory Augmented Neural Networks (MANN) are a class of neural network models that have been recently used to solve complex tasks requiring abstract forms of reasoning and planning (Graves et al., 2016). The wider applicability of such models to real-world problems and data sources is however still left unexplored. The purpose of my research project is threefold: my first aim will be to fill this gap, exploring the scalability potential of MANNs. Furthermore, I will investigate how such models' functioning can be explained to the end user, in line with the most recent social needs regarding applied AI (Murdoch et al., 2019). Finally, I will study how these architectures may be applied to model cognitive abilities, especially in the field of mathematical cognition (Testolin, 2020).