Variational autoencoders (VAEs) are widely used for generative modeling and repre- sentation learning tasks. This thesis presents two novel approaches aimed at enhancing the performance of VAEs through the integration of ...