Deep Neural Networks for Visual Reasoning

This thesis studied how machines learn to reason across vision and language with neural networks. The findings through a series of neural network models with application on visual question answering revealed different pivotal aspects of vision-language interactions to support the reasoning process in machine learning.