Prior and future research can be summarized in three tracks:

1. Non-smooth optimization and non-Newtonian flows. Several efficient optimization methods have been developed. Future work involves learnable optimization solvers using neural networks.

2. Generative models for 3D data. Two efficient approaches using StyleGAN and conditional GAN architectures have been developed. Future work involves conditional generation with specific topological properties. Applications include microstructure synthesis and digital rock physics.

3. Efficient solvers and physics-based machine learning. Efficient reduced order models have been developed for viscoplastic flows. Future work involves the developed of efficient three-dimensional solvers for fluid flow problems