Concept神经-数值专家
Neural-Numerical Mixture of Experts
A design paradigm that combines learnable neural networks with fixed numerical solvers, in which the fixed numerical updates correspond to known discretization schemes and the learnable predictors output the intermediate quantities required by those schemes.
Timeline (last 90 days)
Oct 30
The neural-numerical mixture of experts is a design paradigm that combines learnable neural networks with fixed numerical solvers, where the fixed numerical update contains no trainable parameters and advances the state according to the equation's structured rules, while the learnable predictor outputs the intermediate quantities required by the numerical scheme.
Unverified50%