Examining potential drug-target interactions (DTIs) is a pivotal component of drug discovery and repurposing. Recently, there has been a significant rise in the use of computational techniques to predict DTIs. Nevertheless, previous investigations have predominantly concentrated on assessing either the connections between nodes or the consistency of the network's topological structure in isolation. Such one-sided approaches could severely hinder the accuracy of DTI predictions. In this study, we propose a novel method called TTGCN, which combines heterogeneous graph convolutional neural networks (GCN) and graph attention networks (GAT) to address the task of DTI prediction. TTGCN employs a two-tiered feature learning strategy, utilizing GAT and residual GCN (R-GCN) to extract drug and target embeddings from the diverse network, respectively. These drug and target embeddings are then fused through a mean-pooling layer. Finally, we employ an inductive matrix completion technique to forecast DTIs while preserving the network's node connectivity and topological structure. Our approach demonstrates superior performance in terms of area under the curve and area under the precision-recall curve in experimental comparisons, highlighting its significant advantages in predicting DTIs. Furthermore, case studies provide additional evidence of its ability to identify potential DTIs.

Graph clustering is a significant task in complex network research. Deep graph clustering aims to uncover the potential community structure in graph data using the powerful feature extraction capability of deep learning , garnering much attention in recent decades. However, existing graph clustering methods fall short in fully utilizing available information, particularly in effectively fusing structural and attribute information, as well as utilizing coarse-grained data. Consequently, learned node representations remain limited, leading to suboptimal clustering results . To address these challenges, we propose I nformation- E nhanced D eep G raph C lustering N etwork (IEDGCN) for unsupervised attribute graphs. IEDGCN introduces key components to enhance information utilization and improve clustering performance. Firstly, we design a new higher-order neighborhood-weighted attribute matrix, effectively integrating higher-order neighborhood information with attributes. Secondly, a graph generation model guides the learning of the structural feature space more effectively. Additionally, IEDGCN captures more coarse-grained information by utilizing community and higher-order neighborhood features to refine clustering results. Finally, the proposed method is uniformly guided through a jointly supervised strategy for representation learning and cluster assignment. Experimental results on different benchmark datasets demonstrate the effectiveness of IEDGCN compared to state-of-the-art methods, emphasizing the importance of information enhancement for graph clustering.

Since the data rates in advanced chips dramatically increase, the electrostatic-discharge (ESD) protection circuit at I/O ports causes significant ultrawideband challenges. To address these challenges, an arbitrarily tapped coil (AT-coil) structure with multitaps that enhance bandwidths is proposed; moreover, this structure can also fit into any rectangular layout area to save footprint. A microwave network-assisted analysis approach, which can better consider the nonideal factors in practical applications and thus converge to a better solution than traditional methods, is proposed with regard to the AT-coil. Further, machine learning-assisted optimization (MLAO) is introduced for the designing of AT-coil layouts for the first time, and prior human knowledge of high-speed circuits is maximally utilized to speed up MLAO processes. Thus, the converging stability is significantly improved to escape from a local optimum. Additionally, an automatic synthesis example for an on-chip spiral AT-coil proves the capability. Numerical results show that the proposed approach can achieve the best-high-frequency performance and guarantee ESD protective robustness.

Natural Language Processing (NLP) aims to study various theories and methods that enable effective communication between humans and computers in natural language. One specific technique, known as Keyphrase Extraction (KPE), has achieved significant success in recent years through pre-trained Language Models (LM), particularly BERT and ELMo. Currently, researchers have presented NLP4RE at Requirements Engineering (RE) conferences, contemplating how to leverage the cutting-edge advancements in NLP to achieve the integration of closely related research domains. In the practice of requirements engineering, it is not always assumed that the initial requirements description is complete, which can lead to requirements missing or changes. To address this issue, this paper proposes an unsupervised keyword extraction modeling method. Specifically, using the problem frames model as a case study, this method is integrated into the team's development of an iOS-based Problem Frames (PF) modeling tool in the form of an assisting dropdown list. It is linked with external knowledge bases to predict new keywords. We compare five unsupervised keyword extraction techniques with different principles and evaluate them using samples from the requirements engineering domain. In addition, an eye movement experiment is conducted to further assess the proposed method.

An inductor is one of the basic passive circuit elements that make up integrated circuits. A hybrid knowledge-guided and data-driven technique (HKDT) that combines prior knowledge of on-chip inductors with a machine learning approach is proposed for designing a millimeter-wave on-chip deformable spiral inductor (DSI). First, a DSI, whose shape is no longer limited to a regular polygon and can be adapted to any rectangular region, is presented. Next, an expanded Wheeler formula is proposed to estimate the inductance of the DSI. Then, two approximate expressions with fitting parameters are deduced for the frequency responses of the inductance and quality factor. After that, two prior knowledge-guided Gaussian process regression (GPR) surrogate models are presented for the inductance and quality factor, with which the feature dimensions of the training data can be significantly reduced. Both models can achieve higher prediction accuracies in the frequency range of interest with lower computational complexity than the traditional GPR model with frequency-domain features when they are used for machine learning-assisted optimization (MLAO). Finally, an automatic inductor synthesis method is implemented by using a multibranch MLAO algorithm, and two examples are presented to validate the proposed synthesis method and illustrate its great capabilities. Ultimately, the HKDT can be extended to and applied for the efficient synthesis of other circuit elements with high accuracy.