University of South Carolina

Glucose digital twins in septic intensive care previously refit the model only after forecast error crossed a threshold, a reactive correction. This work reports a proactive workflow: a frozen PatchTST foundation model produces a one-step forecast, and an Actor–Critic reinforcement-learning agent uses the last 30-minute glucose window and recent residual history to apply a correction bounded at ±0.05 to that first step, which is then rolled out recursively over 15 steps. On minute-by-minute continuous glucose monitoring from 10 septic ICU patients, the pretrained agent lowered one-step forecast error in all 10 without patient-specific retraining, and mean absolute percentage error on the source patient's test set fell from 0.49% to 0.34%.

Glucose digital twins in septic intensive care previously refit the model only after forecast error crossed a threshold, a reactive correction. This work reports a proactive workflow: a frozen PatchTST foundation model produces a one-step forecast, and an Actor–Critic reinforcement-learning agent uses the last 30-minute glucose window and recent residual history to apply a correction bounded at ±0.05 to that first step, which is then rolled out recursively over 15 steps. On minute-by-minute continuous glucose monitoring from 10 septic ICU patients, the pretrained agent lowered one-step forecast error in all 10 without patient-specific retraining, and mean absolute percentage error on the source patient's test set fell from 0.49% to 0.34%.

The authors propose LD-GTransNet: building on the multi-layer GTransNet approach, it adds a lift-and-decoupling strategy with preset hidden-layer neuron parameters and separate subnetworks for physical coordinates and an added auxiliary variable, giving a unified global approximation of piecewise smooth solutions without explicit domain decomposition and reducing the formulation to a linear least-squares problem that removes nonlinear training; numerical experiments on 2D and 3D benchmark interface problems show better accuracy and efficiency than existing neural-network and traditional numerical methods, especially for high-contrast coefficients and complex interface geometries.

The authors propose LD-GTransNet: building on the multi-layer GTransNet approach, it adds a lift-and-decoupling strategy with preset hidden-layer neuron parameters and separate subnetworks for physical coordinates and an added auxiliary variable, giving a unified global approximation of piecewise smooth solutions without explicit domain decomposition and reducing the formulation to a linear least-squares problem that removes nonlinear training; numerical experiments on 2D and 3D benchmark interface problems show better accuracy and efficiency than existing neural-network and traditional numerical methods, especially for high-contrast coefficients and complex interface geometries.