Investigating the Performance and Energy Costs of Replicating Band-Split RNN for Music Source Separation
研究通过实现全管道复制并优化BSRNN模型，解决了音乐源分离中的性能和能耗问题。
0 citationsRead paper
Selected work
Representative Papers
Learning Unsteady Aneurysm Hemodynamics with Physics-Informed DeepONets
This study addresses the computational burden of unsteady hemodynamic assessment in three-dimensional abdominal aortic aneurysms by proposing M3PI-DeepONet. This method introduces a novel adaptive architecture featuring hierarchical gating and multi-branch operator networks, which integrates feature aggregation injection with embedded three-dimensional Navier-Stokes constraints to enable accurate flow and pressure field predictions under limited labeled data. Experimental results demonstrate that the model achieves velocity and pressure errors below 4% and 5%, respectively, while delivering a 36-fold inference speedup over traditional computational fluid dynamics simulations. Consequently, M3PI-DeepONet provides an efficient and reliable intelligent tool for real-time clinical diagnosis of aneurysms, effectively bridging the gap between high-fidelity hemodynamic modeling and clinical applicability through physics-informed deep learning.
0 citationsRead paper
Recent publications
Latest Papers
Investigating the Performance and Energy Costs of Replicating Band-Split RNN for Music Source Separation
研究通过实现全管道复制并优化BSRNN模型，解决了音乐源分离中的性能和能耗问题。
0 citationsRead paper
Learning Unsteady Aneurysm Hemodynamics with Physics-Informed DeepONets
This study addresses the computational burden of unsteady hemodynamic assessment in three-dimensional abdominal aortic aneurysms by proposing M3PI-DeepONet. This method introduces a novel adaptive architecture featuring hierarchical gating and multi-branch operator networks, which integrates feature aggregation injection with embedded three-dimensional Navier-Stokes constraints to enable accurate flow and pressure field predictions under limited labeled data. Experimental results demonstrate that the model achieves velocity and pressure errors below 4% and 5%, respectively, while delivering a 36-fold inference speedup over traditional computational fluid dynamics simulations. Consequently, M3PI-DeepONet provides an efficient and reliable intelligent tool for real-time clinical diagnosis of aneurysms, effectively bridging the gap between high-fidelity hemodynamic modeling and clinical applicability through physics-informed deep learning.
0 citationsRead paper