The field is witnessing a significant shift towards the development of innovative neural representation techniques, particularly implicit neural representations (INRs), which offer resolution independence and high memory efficiency. Researchers are exploring the application of INRs in various domains, including cardiovascular anatomies and hemodynamic fields, with promising results. Another area of focus is the integration of hyperbolic geometry in recommendation systems and neural networks, enabling the capture of complex hierarchical structures and dynamic temporal contexts. Additionally, there is a growing interest in developing scalable and compressible state space models, as well as novel dimensionality reduction methods for physics-based data. Noteworthy papers include:

• A study on the accuracy of INRs for cardiovascular anatomies and hemodynamic fields, demonstrating remarkable compression ratios and low absolute errors.
• The proposal of GTR-Mamba, a framework for cross-manifold conditioning and routing in hyperbolic geometry, which consistently outperforms state-of-the-art baseline models in next POI recommendation.
• The introduction of Hankel singular value regularization for highly compressible state space models, leading to fast decay of singular values and thus to compressible models.
• The development of a data-driven dimensionality reduction method for hyperbolic wave dynamics, utilizing a specialized neural network architecture called low rank neural representation (LRNR).
• A structured local learning framework that operates directly on low-rank manifolds, enabling local, parallelizable updates with lower memory requirements and achieving accuracy comparable to that of backpropagation.