Convolutional Neural Networks (CNNs) have been widely used in machine learning tasks. While delivering state-of-the-art accuracy, CNNs are known as both compute- and memory-intensive. This paper presents the SqueezeFlow accelerator architecture that exploits sparsity of CNN models for increased efficiency. Unlike prior accelerators that trade complexity for flexibility, SqueezeFlow exploits concise convolution rules to benefit from the reduction of computation and memory accesses as well as the acceleration of existing dense architectures without intrusive PE modifications. Specifically, SqueezeFlow employs a PT-OS-sparse dataflow that removes the ineffective computations while maintaining the regularity of CNN computations. We present a full design down to the layout at 65 nm, with an area of $4.80 mm^2$ and power of 536.09 mW. The experiments show that SqueezeFlow achieves a speedup of 2.9x on VGG16 compared to the dense architectures, with an area and power overhead of only 8.8% and 15.3%, respectively. On three representative sparse CNNs, SqueezeFlow improves the performance and energy efficiency by 1.8x and 1.5x over the state-of-the-art sparse accelerators.

Convolutional neural network (CNN) accelerators have achieved great success from cloud to edge scenarios. However, given the trend towards even larger and deeper neural network models, it remains a challenging problem to efficiently process these CNNs especially on edge devices with limited energy budget. Accordingly, reducing the energy consumption is of paramount importance for sustainable CNN accelerators. In this paper, we propose AdaPrune, a novel pruning technique that reduces model size and computation to achieve performance improvement and energy savings for CNN accelerators. Unlike previous pruning techniques that sacrifice either computational regularity or accuracy, AdaPrune maintains both by customizing CNN pruning for the underlying accelerators to maximally leverage the sparsity benefits. AdaPrune consists of two techniques: input channel group pruning and output channel group pruning. By analyzing the weight fetching patterns of sparse CNN accelerators, AdaPrune adaptively switches between the two techniques to guarantee that the zeros are evenly distributed in each fetching group. In doing so, the pruned network structure preserves customized computational regularity for the underlying accelerators, thereby boosting the performance and energy efficiency. We evaluate AdaPrune on three sparse CNN accelerators with different spatial tiling strategies. The experimental results show that AdaPrune achieves up to 1.6× performance speedup, and 1.5× energy savings compared to unstructured pruning.

Punjab, the most populous province in Pakistan, is currently facing substantial electricity shortages that are adversely affecting both residential and industrial sectors. To address this issue, the Cholistan Desert presents a promising solution due to its high solar irradiance, making it an ideal location for solar energy production. This study aims to identify the most suitable area for solar photovoltaic (PV) power plants in the Cholistan Desert using Geographic Information System (GIS) and machine learning techniques. The analysis included field survey data encompassing 14 conditioning factors such as geophysical, socio-economic, and resource conditions. Three machine learning models were utilized: Random Forest, XGBoost, and Multilayer Perceptron (MLP). The Random Forest model demonstrated superior performance with an AUC of 0.92, and feature importance was measured through SHAP. The resulting suitability map indicates that Bahawalnagar in the eastern region and Bahawalpur in the central region have 10.50% and 11.06% of their areas classified as having a “high” and “very high” probability for solar PV installation, respectively. For stakeholders in the wind industry, these regions also present potential for wind farm feasibility due to favorable wind conditions and flat terrain. The methodology can be adapted to prioritize wind energy sites by incorporating factors such as land availability, wind direction, and other related factors. Co-locating solar and wind farms in these regions could optimize land use, enhance grid stability, and support Pakistan’s renewable energy targets. Future research integrating real-time solar and wind data could further refine site selection and support multi-source renewable energy planning, providing actionable insights for policymakers and investors.

Partial Differential Equations (PDEs) are widely employed to describe natural phenomena in many science and engineering fields. Many PDEs do not have analytical solutions, hence, numerical methods have become prevalent for approximating PDE solutions. The most widely used numerical method is the Finite Difference Method (FDM), which requires fine grids and high-precision numerical iterations that are both compute- and memory-intensive. PDE-solving accelerators have been proposed in the literature, however, they usually focus on specific types of PDEs with rigid grid sizes which limits their broader applicability. Besides, they rarely provided insight into the optimizations of parallel computing and data accesses for solving PDEs, which hinders further improvements in performance and energy efficiency. This paper presents FDMAX, an elastic accelerator to efficiently support FDM for different types of PDEs with any grid size. FDMAX employs a customized Processing Element (PE) array architecture that maximizes data reuse with minimized interconnection overhead. The PE array can be reconfigured to break into a set of subarrays to adapt to different grid sizes for optimal efficiency. Moreover, the PE array exploits computation and data reuse for increased performance and energy efficiency, and is reconfigurable to support a wide range of PDEs such as elliptic, parabolic, and hyperbolic equations. Evaluated on four well-known PDEs, our simulation results show that FDMAX achieves on average 1189× speedup with 1123× energy reduction over Intel Xeon CPU, and 4.9× speedup with 6.3× energy reduction over NVIDIA RTX3090 GPU, and 2.9× speedup over Alrescha, the state-of-the-art PDE-solving accelerator.

Convolutional Neural Networks (GCNs) have emerged as promising tools for graph-based machine learning applications. Given that GCNs are both compute- and memory-intensive, this constitutes a major challenge for the underlying hardware to efficiently process large-scale GCNs. In this article, we introduce SGCNAX, a scalable GCN accelerator architecture for the high-performance and energy-efficient acceleration of GCNs. Unlike prior GCN accelerators that either employ limited loop optimization techniques, or determine the design variables based on random sampling, we systematically explore the loop optimization techniques for GCN acceleration and propose a flexible GCN dataflow that adapts to different GCN configurations to achieve optimal efficiency. We further propose two hardware-based techniques to address the workload imbalance problem caused by the unbalanced distribution of zeros in GCNs. Specifically, SGCNAX exploits an outer-product-based computation architecture that mitigates the intra-PE (Processing Elements) workload imbalance, and employs a group-and-shuffle approach to mitigate the inter-PE workload imbalance. Simulation results show that SGCNAX performs 9.2×, 1.6× and 1.2× better, and reduces DRAM accesses by a factor of 9.7×, 2.9× and 1.2× compared to HyGCN, AWB-GCN, and GCNAX, respectively.

Sonar imagery is substantially degraded by speckle noise, making the task of despeckling crucial for improving image quality. Self-supervised despeckling methods, represented by blind-spot networks (BSNs), have shown promise in this regard. However, these methods consistently face significant challenges due to the spatial correlation of speckle noise and the inherent information loss within BSNs. In this paper, we introduce SEGSID, a BSN-based, semantic-guided sonar despeckling framework designed to address these challenges. Specifically, the SEGSID framework primarily comprises a Receptive Field Augmentation (RFA) module and a Global Semantic Enhancement (GSE) module. To address the noise spatial correlation, the RFA module is crafted to strategically extract valuable local information while avoiding the exploitation of noise-correlated pixels. Concurrently, the GSE module extracts the global semantic information from entire images and injects it into the extracted local features. This enhances BSNs’ ability to harness more comprehensive image information and compensates for their inherent information loss. Furthermore, to bolster efficiency, we employ knowledge distillation techniques to transfer the expertise from the trained SEGSID into a more streamlined network suitable for broader practical applications. Extensive experiments on three distinct sonar datasets demonstrate that SEGSID outperforms both traditional despeckling methods and state-of-the-art self-supervised despeckling techniques. The implementation is publicly accessible at https://github.com/deng-ai-lab/SEGSID.