Zhenyu Wang

Zhenyu Wang, Shahriar Nirjon

Edge devices, with their widely varying capabilities, support a diverse range of edge AI models. This raises the question: how does an edge model differ from a high-accuracy (base) model for the same task? We introduce XDELTA, a novel explainable AI tool that explains differences between a high-accuracy base model and a computationally efficient but lower-accuracy edge model. To achieve this, we propose a learning-based approach to characterize the model difference, named the DELTA network, which complements the feature representation capability of the edge network in a compact form. To construct DELTA, we propose a sparsity optimization framework that extracts the essence of the base model to ensure compactness and sufficient feature representation capability of DELTA, and implement a negative correlation learning approach to ensure it complements the edge model. We conduct a comprehensive evaluation to test XDELTA's ability to explain model discrepancies, using over 1.2 million images and 24 models, and assessing real-world deployments with six participants. XDELTA excels in explaining differences between base and edge models (arbitrary pairs as well as compressed base models) through geometric and concept-level analysis, proving effective in real-world applications.

Wang Zhenyu, Li Peiyuan, Shi Yongxiang et al.

Full-waveform inversion (FWI) can produce high-resolution subsurface models, yet it remains inherently ill-posed, highly nonlinear, and computationally intensive. Although recent deep learning and numerical acceleration methods have improved speed and scalability, they often rely on single CNN architectures or single neural operators, which struggle to generalize in unknown or complex geological settings and are ineffective at distinguishing diverse geological types. To address these issues, we propose a Structure-Aligned Encoder-Mixture-of-Operators (SA-EMO) architecture for velocity-field inversion under unknown subsurface structures. First, a structure-aligned encoder maps high-dimensional seismic wavefields into a physically consistent latent space, thereby eliminating spatio-temporal mismatch between the waveform and velocity domains, recovering high-frequency components, and enhancing feature generalization. Then, an adaptive routing mechanism selects and fuses multiple neural-operator experts, including spectral, wavelet, multiscale, and local operators, to predict the velocity model. We systematically evaluate our approach on the OpenFWI benchmark and the Marmousi2 dataset. Results show that SA-EMO significantly outperforms traditional CNN or single-operator methods, achieving an average MAE reduction of approximately 58.443% and an improvement in boundary resolution of about 10.308%. Ablation studies further reveal that the structure-aligned encoder, the expert-fusion mechanism, and the routing module each contribute markedly to the performance gains. This work introduces a new paradigm for efficient, scalable, and physically interpretable full-waveform inversion.

Zhenyu Wang, Mahathir Monjur, Shahriar Nirjon

In mmWave-based pose estimation, sparse signals and weak reflections often cause models to infer body joints from statistical priors rather than sensor data. While prior knowledge helps in learning meaningful representations, over-reliance on it degrades performance in downstream tasks like gesture and activity recognition. In this paper, we introduce mmJoints, a framework that augments a pre-trained, black-box mmWave-based 3D pose estimator's output with additional joint descriptors. Rather than mitigating bias, mmJoints makes it explicit by estimating the likelihood of a joint being sensed and the reliability of its predicted location. These descriptors enhance interpretability and improve downstream task accuracy. Through extensive evaluations using over 115,000 signal frames across 13 pose estimation settings, we show that mmJoints estimates descriptors with an error rate below 4.2%. mmJoints also improves joint position accuracy by up to 12.5% and boosts activity recognition by up to 16% over state-of-the-art methods.