Bridging the Gap: Unifying the Training and Evaluation of Neural Network Binary Classifiers On the Discrimination Risk of Mean Aggregation Feature Imputation in Graphs Efficient and Modular Implicit Differentiation NeuroSchedule: A Novel Effective GNN-based Scheduling Method for High-level Synthesis