The field of thermal management and neural networks is rapidly evolving, with a focus on optimizing heat transfer and developing more efficient neural network architectures. Recent research has explored the use of unit-cell shape optimization and adjoint-based optimization methods to enhance heat transfer in compact cooling solutions and heat exchangers. Additionally, innovative neural network architectures such as hybrid real- and complex-valued neural networks and physics-informed Kolmogorov-Arnold Networks (PIKANs) have been proposed to improve the accuracy and efficiency of neural network-based solutions. Noteworthy papers in this area include the proposal of a non-parametric B-spline decoupling algorithm for representing multivariate functions and the development of a surrogate-assisted simulated annealing algorithm for fast thermal-aware chiplet placement. The DeepOHeat-v1 framework, which integrates Kolmogorov-Arnold Networks with learnable activation functions, has also shown promising results in accelerating thermal simulation and optimization. These advancements have the potential to significantly impact various fields, including engineering, physics, and computer science, and are expected to continue shaping the direction of research in thermal management and neural networks.
Advances in Thermal Management and Neural Networks
Sources
A Unit-Cell Shape Optimization Approach for Maximizing Heat Transfer in Periodic Fin Arrays at Constant Solid Temperature
Non-parametric B-spline decoupling of multivariate functions
Optimal Sizing and Material Choice for Additively Manufactured Compact Plate Heat Exchangers
Hybrid Real- and Complex-valued Neural Network Architecture
Fast Thermal-Aware Chiplet Placement Assisted by Surrogate
A Machine Learning and Finite Element Framework for Inverse Elliptic PDEs via Dirichlet-to-Neumann Mapping
DeepOHeat-v1: Efficient Operator Learning for Fast and Trustworthy Thermal Simulation and Optimization in 3D-IC Design
Truncated Huber Penalty for Sparse Signal Recovery with Convergence Analysis
Solving the fully nonlinear Monge-Amp\`ere equation using the Legendre-Kolmogorov-Arnold Network method
Multi-level Neural Networks for high-dimensional parametric obstacle problems
Adversarial KA
Eikonal boundary condition for level set method
Orthogonal Matching Pursuit based Reconstruction for Modulo Hysteresis Operators
Physics-informed KAN PointNet: Deep learning for simultaneous solutions to inverse problems in incompressible flow on numerous irregular geometries
Neural Network Enhanced Polyconvexification of Isotropic Energy Densities in Computational Mechanics
Aplicando diferencias finitas para resolver ecuaciones y sistemas de ecuaciones diferenciales parciales sobre dominios planos irregulares simplemente conexos y no conexos
Optimality of Gradient-MUSIC for Spectral Estimation
Compound and Parallel Modes of Tropical Convolutional Neural Networks