The deployment of Deep Neural Networks (DNNs) on edge devices is hindered by the substantial gap between performance requirements and available computational power. While recent research has made significant strides in developing pruning methods to build a sparse network for reducing the computing overhead of DNNs, there remains considerable accuracy loss, especially at high pruning ratios. We find that the architectures designed for dense networks by differentiable architecture search methods are ineffective when pruning mechanisms are applied to them. The main reason is that the current methods do not support sparse architectures in their search space and use a search objective that is made for dense networks and does not focus on sparsity. This paper proposes a new method to search for sparsity-friendly neural architectures. It is done by adding two new sparse operations to the search space and modifying the search objective. We propose two novel parametric SparseConv and SparseLinear operations in order to expand the search space to include sparse operations. In particular, these operations make a flexible search space due to using sparse parametric versions of linear and convolution operations. The proposed search objective lets us train the architecture based on the sparsity of the search space operations. Quantitative analyses demonstrate that architectures found through DASS outperform those used in the state-of-the-art sparse networks on the CIFAR-10 and ImageNet datasets. In terms of performance and hardware effectiveness, DASS increases the accuracy of the sparse version of MobileNet-v2 from 73.44% to 81.35% (+7.91% improvement) with a 3.87× faster inference time.

Federated learning (FL) offers a privacy-preserving solution for single-image super-resolution (SR) on sensitive satellite imagery, but its performance is often hindered by simplistic data models. Existing methods that rely on simple bicubic downsampling fail to capture the complex, client-specific degradations found in real-world satellite data, creating a significant domain gap. To address this, we propose a novel strategy, Heterogeneous Realistic Degradation (HeRD), which models data heterogeneity by generating realistic low-resolution images based on the unique, device-locked characteristics of different satellites. Unlike conventional approaches, HeRD systematically applies diverse, anisotropic degradations to enable fine-grained control over non-Independent and Identically Distributed (non-IID) conditions. Our extensive evaluations demonstrate the robustness of FL when trained with HeRD. The proposed federated pipeline outperforms traditional bicubic-based setups by over 0.5 dB in PSNR. Notably, even in highly heterogeneous environments, our approach achieves performance within just 0.2–0.4 dB of a fully centralized training model. These findings confirm that HeRD provides a viable, high-performance, and privacy-preserving alternative for super-resolving distributed satellite imagery where data sovereignty and disparate hardware characteristics are paramount.1

Designing noise-robust parameterized quantum circuits (PQCs) is a central challenge in the noisy intermediate-scale quantum (NISQ) regime. Existing quantum architecture search methods rely on training large SuperCircuits and evaluating SubCircuits under noisy execution, resulting in high computational cost and architecture assessments that depend on task-specific optimization and device noise. In this work, we propose a training-free quantum architecture search framework based on information-theoretic expressibility measures rather than performance-based estimators. We empirically show that noise-free KL-divergence-based expressibility exhibits a consistent monotonic association with noisy task loss across diverse circuit architectures and realistic hardware noise models. Leveraging this relationship, we introduce an expressibility-guided evolutionary search that requires neither SuperCircuit training nor noisy execution during the search phase. Since expressibility is evaluated independently of hardware noise, the method is inherently device-agnostic, enabling architectures to be reused across multiple quantum devices without re-running the search. Experiments using IBM-derived Qiskit noise models demonstrate that the proposed approach achieves competitive performance compared to SuperCircuit-based baselines, while substantially reducing computational cost. These results establish expressibility as an effective information-theoretic surrogate for ranking PQC architectures under realistic noise.

Convolutional neural networks used in safety-critical systems must remain reliable under hardware-induced faults. Activation clipping is an effective fault-mitigation technique, yet existing approaches face a fundamental trade-off: layer-wise clipping provides limited resilience, whereas neuron-wise clipping introduces substantial memory overhead. This study proposes a hybrid clipped Rectified Linear Unit activation function that combines both strategies by applying neuron-wise clipping exclusively to the final layer and layer-wise clipping elsewhere. To further improve reliability, we introduce a progressive training strategy that optimizes the clipping thresholds in a layer-by-layer manner using knowledge distillation. This progressive optimization enables tighter threshold selection at each layer, improving fault tolerance while maintaining low memory costs. Experimental results under high error rates show that the proposed approach significantly enhances reliability compared with existing methods, achieving higher accuracy while substantially reducing memory overhead.