Deep learning has been extensively employed as a powerful function approximator for modeling physics-based problems described by partial differential equations (PDEs). Despite their popularity, standard deep learning models often demand prohibitively large computational resources and yield limited accuracy when scaling to large-scale, high-dimensional physical problems. Their black-box nature further hinders their application in industrial problems where interpretability and high precision are critical. To overcome these challenges, this paper introduces Interpolating Neural Network-Tensor Decomposition (INN-TD), a scalable and interpretable framework that has the merits of both machine learning and finite element methods for modeling large-scale physical systems. By integrating locally supported interpolation functions from finite element into the network architecture, INN-TD achieves a sparse learning structure with enhanced accuracy, faster training/solving speed, and reduced memory footprint. This makes it particularly effective for tackling large-scale high-dimensional parametric PDEs in training, solving, and inverse optimization tasks in physical problems where high precision is required.

This paper presents a novel Double-Balanced power amplifier (PA) architecture with an intrinsic load isolation. Derived from the generic load modulated balanced amplifier (LMBA), by designing the single-ended control amplifier (CA) as another balanced PA, the Pseudo-Doherty load-modulated double-balanced amplifier (PD-LMDBA) can inherit the intrinsic load-mismatch tolerance of balanced amplifier without any reconfiguration and load-impedance sensing. Theoretical analysis reveals that both the control amplifier (CA, as carrier) and primary balanced amplifier (BA, as peaking) exhibit complementary load modulation trajectories for their sub-amplifiers (CA1 and CA2, BA1 and BA2) under mismatch. This allows the PA to inherit the intrinsic load insensitivity from the generic quadrature-balanced amplifier and sustain nearly constant performance against arbitrary load variations. A prototype is implemented at 2.1 GHz, achieving 76.2% efficiency at peak power and 69.5% at 10-dB OBO with matched load. Under a$2:1$voltage standing wave ratio (VSWR) of load mismatch, an efficiency up to 72.5% at peak power and 64.1% at 10-dB OBO are measured. In modulated evaluation with a 20-MHz OFDM signal, the PA maintains linearity against$2:1$VSWR, with 2.1% of error vector magnitude (EVM) and down to -39.5 dB adjacent channel power ratio (ACPR), closely approximating the matched condition.