Current technology trends have led to the growing impact of both inter-die and intra-die process variations on circuit performance. While it is imperative model parameter for sub-100nm technologies produce an upper bound prediction timing, equally important consider correlation these be useful. In this paper we present efficient block-based statistical static timing analysis algorithm that can account correlations from parameters re-converging paths. The also accommodate dominant...

Water environmental Internet of Things (IoT) system, which is composed multiple monitoring points equipped with various water quality IoT devices, provides the possibility for accurate prediction. In same area, flows and exchanges between points, resulting in an adjacency effect information. However, traditional prediction methods only use information one point, ignoring nearby points. this paper, we propose a method based on multi-source transfer learning order to effectively improve...

Due to process scaling, variability in process, voltage, and temperature (PVT) parameters leads a significant parametric yield loss, thus impacts the optimization for circuit designs seriously. Previous algorithms are limited optimizing either power or timing separately, without combining them together simultaneous optimization. However, neglecting negative correlation between performance metrics, such as measurements, will bring on accuracy loss. This paper suggests an efficient...

With the integration of cyber-physical system and cloud computing, virtual machine (VM) placement has been great importance to performance (CPCS). This paper proposes a stochastic VM algorithm that takes into account uncertainty resource requirements while minimizing total energy consumption in CPCS. Different from existing approaches use deterministic values represent demands, proposed models as random variables, further formulates uncertainty-affected problem optimization model. The...

With the successful integration of contrastive learning and graph neural networks, (GCL) has demonstrated superior performance in representation learning. Majority earlier works tend to utilize dual-view frameworks. However, they require high computational costs; additionally, we observed that can hardly obtain robust result as training processes swing between two important metrics: alignment uniformity. We address these problems by designing a novel single-view paradigm called Light...

Training a semantic segmentation model requires large densely-annotated image datasets that are costly to obtain. Once the training is done, it also difficult add new object categories such models. In this paper, we tackle few-shot problem, which aims perform task on unseen merely based one or few support example(s). The key solving problem lies in effectively utilizing information from examples separate target objects background query image. While existing methods typically generate...

Organizing training samples in a meaningful order is beneficial for accelerating the convergence rate and enhancing recognition performance CNN model. However, achieving reasonable sample ranking fine-grained datasets very challenging because intra inter class relation those opposite to that public datasets. In this paper, we propose general framework progressive of models. particular, first formulate subset selection as group ranking-oriented submodular optimization problem, where...

The problem of graph classification in neural networks has received extensive research. In the tasks, pooling is one important research directions, which aims to reduce dimensionality representation and obtain prediction label for entire graph. However, due fixed compression quotas node discarding strategies, commonly adopted hierarchical methods often suffer structure destruction information loss. And their model training will be time-consuming. To address these issues, we suggest a method...

The link prediction problem has been extensively studied in graph neural networks. However, there are still many problems to be solved prediction. when generating node embeddings, using some methods unsupervised learning can lead inefficiency, lack of accuracy, and failure reflect the structural features network, which a significant impact on accuracy later predictions. Therefore, we incorporate random walk strategy generate initial embeddings model, improves richness quality embeddings....

Due to the aggressive shrinking of integrated circuit (IC) fabrication technology, variability in design parameters leads significant yield loss. Therefore, efficient parametric prediction has been a critical task for today’s ICs, especially when multiple performances are considered simultaneously. However, most previous works have failed take into account uncertainty performance-relevant structures. Neglecting effects will tend result predictive accuracy To avoid issue, this paper proposes...

Self-taught learning is a promising technology that utilizes easily accessible label-free images to narrow the performance gap in few-shot learning. During data collection, it naturally comes along with numerous partially occluded samples miss discriminative information. However, few works have seriously studied impact of on self-taught In this paper, we propose cross combination oriented auto-encoder for The proposed composed multi-scale Residual Connection Unions (RCU). innovation RCU lies...

In recent years, Transformer models have been widely developed in various fields, such as natural language processing (NLP) and computer vision (CV). However, its application graph data structures is still limited. It crucial to develop an improved model for data. Considering the weakness that original Graph (GT) difficult capture location information of nodes significantly, this paper combines node positional embedding each layer's output. Then, order select more efficient neighbor nodes, a...

Graph neural networks have attracted more and attention in the task of learning node representations. Currently, most graph are usually applied to assortative graphs. They can't perform well disassortative graphs, where adjacent nodes tend different class labels. However, previous studies shown that it is effective for gather information from their neighbors not only graphs but also Based on this insight, paper proposes a novel Component-level Attention Adaptive Convolutional Network...

Graph neural networks (GNNs) have been widely used to process graph data. However, the ability of most traditional models integrate node features and topology is not ideal. In this case, Adaptive Multi-channel Convolution Networks (AMGCN) model proposed redesign structure. It can greatly improve fusion in complex graphs with rich information. Nevertheless, cannot solve problems over-smoothness non-robustness as depth increases. To address issue, we propose A Random Spectral Network (RMSGCN)....

For graph-based semi-supervised learning, Graph Convolutional Networks (GCNs) and their variants has shown outstanding results gained wide attention. Several works on analyzing GCNs from the perspective of spectral graph theory shows that function low-pass filtering certain learning tasks, which enables us to have a deeper understanding GCN. However, GCN achieves effect through layers matrix multiplication, so one cannot flexibly control process. In this paper, we propose preprocess...