While integrated sensing and communication (ISAC) beamforming has become a research hotspot recently, most proposed approaches remain scenario-specific and suffer from limited generalizability. In this paper, we propose a flexible beamforming design framework for multiple-input multiple-output ISAC (MIMO-ISAC) systems. To characterize the optimal performance tradeoff between the sensing and communication (S&C), the objective is to maximize the weighted sum of S&C mutual information (MI). An integrated fractional programming (IFP) framework is first proposed, from which a semi-closed-form solution for the optimal beamformer is derived. The proposed framework is further extended to several representative ISAC scenarios, including fairness-aware tradeoffs, imperfect channel state information (CSI), dynamic radar cross-section (RCS) variations, and massive MIMO systems. By incorporating advanced optimization techniques, the IFP framework effectively addresses the challenges arising in these scenarios with only minor modifications while still yielding semi-closed-form solutions. Moreover, we show that the IFP framework is closely related to the projected gradient descent (PGD) method, and its computational efficiency can be further enhanced through gradient acceleration techniques. Simulation results demonstrate that the IFP framework and its variants consistently outperform benchmark schemes across diverse system settings. The proposed low-complexity and accelerated schemes reduce computation time by approximately 65% and 89%, respectively, with only marginal performance degradation. These results highlight the flexibility and generality of the proposed IFP framework for ISAC beamforming design and are expected to provide valuable insights for the development of future 6G systems.

Compatibility between semantic communication and traditional mobile communication systems remains a significant challenge. Therefore, we propose a cross-layer encrypted semantic communication (CLESC) framework for panoramic video transmission, incorporating feature extraction, encoding, encryption, cyclic redundancy check (CRC), and retransmission processes to achieve compatibility between semantic communication and traditional communication systems. Additionally, we propose an adaptive cross-layer transmission mechanism that dynamically adjusts CRC, channel coding, and retransmission schemes based on the importance of semantic information. This mechanism ensures that important information is prioritized under poor transmission conditions. To verify the aforementioned framework, we design an end-to-end adaptive panoramic video semantic transmission (APVST) network that leverages a deep joint source-channel coding (JSCC) structure and attention mechanism, integrated with a latitude adaptive module that facilitates adaptive semantic feature extraction and variable-length encoding of panoramic videos. Simulation results demonstrate that the proposed CLESC framework effectively achieves compatibility and adaptability between semantic and traditional communication systems, significantly enhancing channel robustness. Compared to traditional and artificial intelligence (AI)-based video source coding transmission schemes, our proposed CLESC achieves superior transmission performance under low signal-to-noise ratio (SNR) conditions.

In order to satisfy the requirements for service transformation and upgrading toward industrial digitization, networking, and intelligence, sixth generation-enabled industrial Internet of Things (IIoT) imposes new requirements on deterministic delay. However, the existing best-effort communication networks increase the uncertainty of transmission, making it difficult for users to ensure deterministic delay performance. In this article, we propose a deterministic delay guarantee strategy (DDGS) under reconfigurable intelligent surface (RIS)-assisted communication networks to ensure network performance in IIoT scenarios. In particular, we utilize stochastic network calculus (SNCs) to derive the probability that the delay falls within a specific time window, characterizing the probabilistic bounds of deterministic delay. Then, we explore the relationship between delay determinacy and wireless resources by jointly optimizing the transmit power, the channel blocklength allocation, and the phase-shift matrix at the RIS to maximize delay determinacy. Based on the interdependence of action choices among users and past experience, this article proposes a performance guarantee parameterized deep Q-network (PG-PDQN) algorithm to solve the complex problem containing a mixture of discrete and continuous action spaces. Simulation results show that the DDGS strategy significantly improves the delay determinacy compared to other strategies, and the PG-PDQN algorithm has good convergence, thus effectively improving the network performance.

Semantic communication (SemCom) is regarded as a promising and revolutionary technology in 6G, aiming to transcend the constraints of “Shannon’s trap” by filtering out redundant information and extracting the core of effective data. Compared to traditional communication paradigms, SemCom offers several notable advantages, such as reducing the burden on data transmission, enhancing network management efficiency, and optimizing resource allocation. Numerous researchers have extensively explored SemCom from various perspectives, including network architecture, theoretical analysis, potential technologies, and future applications. However, as SemCom continues to evolve, a multitude of security and privacy concerns have arisen, posing threats to the confidentiality, integrity, and availability of SemCom systems. This paper presents a comprehensive survey of the technologies that can be utilized to secure SemCom. Firstly, we elaborate on the entire life cycle of SemCom, which includes the model training, model transfer, and semantic information transmission phases. Then, we identify the security and privacy issues that emerge during these three stages. Furthermore, we summarize the techniques available to mitigate these security and privacy threats, including data cleaning, robust learning, defensive strategies against backdoor attacks, adversarial training, differential privacy, cryptography, blockchain technology, model compression, and physical-layer security. Lastly, this paper outlines future research directions to guide researchers in related fields.

In immersive communication, delivering real-time, high-resolution 360-degree panoramic videos imposes extremely high demands on network performance. In this paper, we propose a rate splitting multiple access (RSMA)-enabled adaptive panoramic video semantic transmission (APVST) framework. Specifically, APVST is built based on the deep joint source-channel coding (JSCC) structure and achieves adaptive semantic extraction and variable-length coding of panoramic frames. Additionally, APVST employs an entropy model and a latitude adaptive module to jointly achieve rate control, and utilizes a weight attention module to enhance the panoramic video quality. Given the overlapping field of view (FoV) when users watch panoramic videos, RSMA is integrated into the semantic transmission to further improve system efficiency. Therefore, we introduce an RSMA-enabled semantic stream transmission scheme, and formulate a joint optimization problem for latency and video quality by optimizing power, common rate, and channel bandwidth allocation ratios, aiming to maximize the users’ quality of service (QoS). To address this problem, we develop a deep reinforcement learning (DRL) approach based on the proximal policy optimization (PPO) algorithm, which integrates semantic-level FoV information to effectively adapt to dynamically changing environments. Simulation results indicate that our proposed APVST reduces bandwidth consumption by 20% compared to semantic video transmission schemes and 45% compared to traditional ones. Furthermore, our research validates the effectiveness of RSMA in panoramic video semantic transmission, demonstrating QoS improvements of up to 20% compared to other multiple access schemes.

Abstract Adolescent idiopathic scoliosis is a significant health concern, ranked as the third most prevalent issue among adolescents after obesity and myopia. Traditional screening methods rely on the use of complex and expensive measuring instruments and expert physicians to interpret X-ray images. These methods can be both time-consuming and inaccessible for widespread screening efforts. To address these challenges, we propose a standardized protocol for the collection of scoliosis gait dataset. This protocol enables the systematic capture of relevant gait characteristics associated with scoliosis, leading to the creation of a comprehensive, annotated dataset tailored for research and diagnostic purposes. Leveraging this dataset, we developed an effective deep learning algorithm based on graph convolutional networks, which outperforms traditional CNN by effectively modeling the complex spatial and temporal dynamics of human gait and posture, leveraging skeletal structure as a graph for more accurate and robust scoliosis screening. We also explored various optimization strategies to enhance the model’s accuracy and efficiency, ensuring robust performance across diverse scenarios. Our innovative approach allows for the rapid and non-invasive recognition of scoliosis. This method is not only scalable but also eliminates the need for specialized equipment or extensive medical expertise, making it ideal for large-scale screening initiatives. By improving the accessibility and efficiency of scoliosis detection, our approach has the potential to facilitate early intervention.

In this paper, we investigate the computation offloading problem in a distributed reconfigurable intelligent surface (RIS)-aided cell-free network, where users offload computing-intensive tasks to their associated base stations with the aid of multiple RISs. To minimize the long-term energy consumption of all users under the constraints of latency tolerance, we formulate a long-term non-convex problem by jointly optimizing the offloading strategy, transmit power at users, reflection matrix at the RIS, receive beamforming at the BS, and CPU resources at both users and the server. To solve this intractable time-varying problem with multiple coupling variables, we propose a two-layer distributed proximal policy optimization (DPPO) algorithm for solving the problem with a high-dimensional decision-making space. Simulation results show that the proposed algorithm effectively reduces the long-term energy consumption of all users while completing the computing task within a given time limit.

In this paper, we investigate a distributed semantic communication system where each device utilizes the semantic model for intelligent tasks. Due to the challenges associated with heterogeneity of federated learning framework, we propose a semantic feature-based learning framework (SFLF) for task-oriented semantic communication. Additionally, we define the global task score to evaluate the task execution performance of whole devices covered by the edge server and formulate a global task score maximization problem under energy constraints. For such framework, the edge server aggregates the semantic features extracted from devices and broadcasts global semantic feature to devices. The devices utilize their local data and the received global semantic feature to train their semantic models by the task loss function and semantic loss function. Besides, since the practical distributed system is usually energy-limited, we propose a distributed proximal policy optimization (DPPO)-based scheduling algorithm to adjust energy consumption in real-time during the training process. Simulation results demonstrate that the proposed SFLF with the DPPO-based device scheduling algorithm outperforms the existing schemes.

In this paper, we propose an adaptive panoramic video semantic transmission (APVST) network built on the deep joint source-channel coding (Deep JSCC) structure for the efficient end-to-end transmission of panoramic videos. The proposed APVST network can adaptively extract semantic features of panoramic frames and achieve semantic feature encoding. To achieve high spectral efficiency and save bandwidth, we propose a transmission rate control mechanism for the APVST via the entropy model and the latitude adaptive model. Besides, we take weighted-to-spherically-uniform peak signal-to-noise ratio (WS-PSNR) and weighted-to-spherically-uniform structural similarity (WS-SSIM) as distortion evaluation metrics, and propose the weight attention module to fuse the weights with the semantic features to achieve better quality of immersive experiences. Finally, we evaluate our proposed scheme on a panoramic video dataset containing 208 panoramic videos. The simulation results show that the APVST can save up to 20% and 50% on channel bandwidth cost compared with other semantic communication-based and traditional video transmission schemes.

We investigate the joint communication and computation problem in non-orthogonal multiple access based mobile edge computation networks for massive intelligent machine type communication. We model the whole task offloading process as a double tandem queues model and formulate an optimization problem to maximize effective energy efficiency while guaranteeing the End-to-End delay violation probability. To solve this problem, we propose a Joint Transmission Power allocation and Computation Resources allocation (JTPCR) algorithm. Specifically, we first minimize the task processing delay to obtain the optimal computation resource constrained by total computation resources and maximum tolerable delay. In addition, we exploit the Dinkelbach method to solve the fractional programming problem when optimizing the transmission power. We introduce an auxiliary variable and obtain the lower bounding concave approximation of channel capacity through a path-following method. Finally, we propose the alternating direction method of the multipliers to obtain the optimal transmission power. Simulation results show that the proposed JTPCR algorithm outperforms the comparison schemes under both scenarios: finite transmission blocklength and infinite transmission blocklength.

This article proposes a secrecy and energy-efficient device-edge co-inference scheme for resource-constrained Artificial Intelligence of Things (AIoT) devices with physical layer security assistance. Our approach leverages split learning, where the AIoT device executes the initial part of the AI model, and the mobile edge computing server (MECs) computes the remainder, reducing energy consumption (EC) and inference delay. We measure secrecy capacity under the finite blocklength regime to address the vulnerability of intermediate feature data (IFD) to eavesdropping over wireless channels and its short block length characteristics. The objective is to minimize the average EC of the device-edge co-inference by jointly optimizing deep neural network (DNN) model partitioning and resource allocation. We formulate a distributed reinforcement learning-based joint DNN model partitioning and resource allocation (DRPA) algorithm, which uses knowledge-based reinforcement learning for optimal DNN partitioning and a convex optimization approach for resource allocation. Simulation results demonstrate that the DRPA algorithm achieves near-optimal performance, closely matching the results of exhaustive search methods.

Symbiotic radio (SR) is a promising technology to support ultralow-power or even zero-power Internet of Things (IoT) devices in the sixth-generation mobile networks. In this article, we propose an energy-aware symbiotic transmission in a reconfigurable intelligent surface (RIS) enhanced SR system, in which an IoT network embeds its own data passively over cellular downlink signals by backscattering. The base station (BS) serves multiple cellular users (CUs) through time division multiple access (TDMA) and each IoT device is associated with one CU. We formulate the BS’s energy minimization problem subject to the constraints of the minimum amounts of transmission bits required by IoT devices and CUs. The user association, the active transmit beamforming at the BS, the passive reflecting beamforming at the RIS, and the frame division policy are jointly optimized. The formulated problem is a mixed integer nonlinear programming (MINLP) problem, which is NP-hard and nonconvex. We decouple the problem and solve the subproblems alternatively. First, we design a many-to-one swap-matching-based algorithm to solve the user association subproblem. Then, we develop a joint cooperative beamforming and time allocation optimization algorithm based on the alternative optimization (AO) and semidefinite relaxation (SDR) techniques. Simulation results show that the proposed joint user association and cooperative beamforming algorithm brings significant performance gain in reducing the energy consumption of the BS with fast convergence speed compared with other schemes.

This paper investigates the age of information (AoI) and energy efficiency of wireless powered industrial Internet of Everything (IIoE) network, where multiple low-power IIoE devices (IIoEDs) are wirelessly charged by a hybrid access point (HAP) to transmit their sensing information to the control nodes. To enhance the system’s information timeliness with high energy efficiency, we define a novel performance metric, i.e., age-energy efficiency (AEE), which depicts the achievable AoI gain per unit energy consumption. Then, an optimization problem is formulated to maximize the system long-term AEE by jointly optimizing the IIoEDs scheduling and the HAP’s transmit power. Due to the non-convexity of the formulated problem and the intractable challenges with discrete binary variables, we first model the problem as a two-stage discrete-time Markov decision process (MDP) with carefully designed state spaces, action spaces, and reward functions. We then propose a deep reinforcement learning (DRL)-based approach to find the effective scheduling strategy and transmit power. To improve the accuracy of the learned policy, we design a dual-layer deep Q-network (DLDQN) algorithm with fast convergence. Simulation results show that our proposed DLDQN algorithm can improve the AEE by at least 25% when the number of IIoEDs exceeds 50 compared with benchmarks. Moreover, with the proposed DLDQN algorithm, the system long-term AEE can be improved with the increase of the number of IIoEDs.

In a multi-user system, system resources should be allocated to different users. In traditional communication systems, system resources generally include time, frequency, space, and power, so multiple access technologies such as time division multiple access (TDMA), frequency division multiple access (FDMA), space division multiple access (SDMA), code division multiple access (CDMA), and non-orthogonal multiple access (NOMA) are widely used. In semantic communication, which is considered a new paradigm of the next-generation communication system, we extract high-dimensional features from signal sources in a model-based artificial intelligence approach from a semantic perspective and construct a model information space for signal sources and channel features. From the high-dimensional semantic space, we excavate the shared and personalized information of semantic information and propose a novel multiple access technology, named model division multiple access (MDMA), which is based on the resource of the semantic domain. From the perspective of information theory, we prove that MDMA can attain more performance gains than traditional multiple access technologies. Simulation results show that MDMA saves more bandwidth resources than traditional multiple access technologies, and that MDMA has at least a 5-dB advantage over NOMA in the additive white Gaussian noise (AWGN) channel under the low signal-to-noise (SNR) condition.

Leveraging computational methods to generate small molecules with desired properties has been an active research area in the drug discovery field. Towards real-world applications, however, efficient generation of molecules that satisfy multiple property requirements simultaneously remains a key challenge. In this paper, we tackle this challenge using a search-based approach and propose a simple yet effective framework called MolSearch for multi-objective molecular generation (optimization).We show that given proper design and sufficient domain information, search-based methods can achieve performance comparable or even better than deep learning methods while being computationally efficient. Such efficiency enables massive exploration of chemical space given constrained computational resources. In particular, MolSearch starts with existing molecules and uses a two-stage search strategy to gradually modify them into new ones, based on transformation rules derived systematically and exhaustively from large compound libraries. We evaluate MolSearch in multiple benchmark generation settings and demonstrate its effectiveness and efficiency.

As a distributed machine learning paradigm, federated learning (FL) has been regarded as a promising candidate to preserve user privacy in Internet of Things (IoT) networks. Leveraging the waveform superposition property of wireless channels, over-the-air FL (AirFL) achieves fast model aggregation by integrating communication and computation via concurrent analog transmissions. To support sustainable AirFL among energy-constrained IoT devices, we consider that the base station (BS) adopts simultaneous wireless information and power transfer (SWIPT) to distribute global model and charge local devices in each communication round. To maximize the long-term energy efficiency (EE) of AirFL, we investigate a resource allocation problem by jointly optimizing the time division, transceiver beamforming, and power splitting in SWIPT-enabled IoT networks. Considering such multiple closely-coupled continuous valuables, we propose a deep reinforcement learning (DRL) algorithm based on twin delayed deep deterministic (TD3) policy to smartly make downlink and uplink communication strategies with the coordination between the BS and devices. Simulation results show that the proposed TD3 algorithm obtains about 41% EE improvement compared to traditional optimization method and other DRL algorithms.

Training deep neural network models in the presence of corrupted supervision is challenging as the corrupted data points may significantly impact generalization performance. To alleviate this problem, we present an efficient robust algorithm that achieves strong guarantees without any assumption on the type of corruption and provides a unified framework for both classification and regression problems. Unlike many existing approaches that quantify the quality of the data points (e.g., based on their individual loss values), and filter them accordingly, the proposed algorithm focuses on controlling the collective impact of data points on the average gradient. Even when a corrupted data point failed to be excluded by our algorithm, the data point will have a very limited impact on the overall loss, as compared with state-of-the-art filtering methods based on loss values. Extensive experiments on multiple benchmark datasets have demonstrated the robustness of our algorithm under different types of corruption. Our code is available at \url{https://github.com/illidanlab/PRL}.

Recent years have seen a rapid growth of utilizing graph neural networks (GNNs) in the biomedical domain for tackling drug-related problems. However, like any other deep architectures, GNNs are data hungry. While requiring labels in real world is often expensive, pretraining GNNs in an unsupervised manner has been actively explored. Among them, graph contrastive learning, by maximizing the mutual information between paired graph augmentations, has been shown to be effective on various downstream tasks. However, the current graph contrastive learning framework has two limitations. First, the augmentations are designed for general graphs and thus may not be suitable or powerful enough for certain domains. Second, the contrastive scheme only learns representations that are invariant to local perturbations and thus does not consider the global structure of the dataset, which may also be useful for downstream tasks. In this paper, we study graph contrastive learning designed specifically for the biomedical domain, where molecular graphs are present. We propose a novel framework called MoCL, which utilizes domain knowledge at both local- and global-level to assist representation learning. The local-level domain knowledge guides the augmentation process such that variation is introduced without changing graph semantics. The global-level knowledge encodes the similarity information between graphs in the entire dataset and helps to learn representations with richer semantics. The entire model is learned through a double contrast objective. We evaluate MoCL on various molecular datasets under both linear and semi-supervised settings and results show that MoCL achieves state-of-the-art performance.

The integration of mobile-edge computing (MEC) and energy harvesting (EH) can potentially improve the network performances and prolong the battery life of the device. In this article, we study the resource management problem in the device-to-device (D2D)-aided wireless powered MEC networks where one device can forward or execute computation data for other devices with its resources. Our problem seeks to optimize the computation offloading strategy, transmission power, energy transmit power, as well as CPU speed to maximize the long-term utility energy efficiency (UEE). UEE is defined as the achieved computation data per unit energy. Since the formulated problem is in fractional form and hard to solve, we employ the Dinkelbach algorithm to transform the problem into a parametric subtractive form. Furthermore, considering that the formulated problem is time varying and stochastic due to the dynamic task arrival rate and battery level, we transform the long-term problem into deterministic drift-plus-penalty subproblems for each time slot by introducing virtual queues and adopting the Lyapunov optimization theory. The proposed scheme can balance the optimal UEE and stable data queue by introducing the control parameter$V$. Theoretically, we reveal the tradeoff between the UEE and stable queue length for wireless powered MEC systems as$[O(1/V), O(V)]$. Finally, the simulations illustrate the efficiency of the proposed scheme compared with the existed work in terms of the UEE, stable queue length, and battery level.

Thanks to the inherent characteristics of flexible mobility and autonomous operation, unmanned aerial vehicles (UAVs) will inevitably be integrated into 5G/B5G cellular networks to assist remote sensing for real-time assessment and monitoring applications. Most existing UAV-assisted data collection schemes focus on optimizing energy consumption and data collection throughput, which overlook the temporal value of collected data. In this article, we employ Age of Information (AoI) as a performance metric to quantify the temporal correlation among data packets consecutively sampled by the Internet of Things (IoT) devices, and investigate an AoI-energy-aware data collection scheme for UAV-assisted IoT networks. We aim to minimize the weighted sum of expected average AoI, propulsion energy of UAV, and the transmission energy at IoT devices, by jointly optimizing the UAV flight speed, hovering locations, and bandwidth allocation for data collection. Considering the system dynamics, the optimization problem is modeled as a Markov decision process. To cope with the multidimensional action space, we develop a twin-delayed deep deterministic (TD3) policy gradient-based UAV trajectory planning algorithm (TD3-AUTP) by introducing the deep neural network (DNN) for feature extraction. Through simulation results, we demonstrate that our proposed scheme outperforms the deep$Q$-network and actor–critic-based algorithms in terms of achievable AoI and energy efficiency.

In the era of the Internet of Things (IoT), reinforcement learning (RL)-based techniques are promising candidates to handle the intrusion response through the interaction between the IoT device and its environment. Given a large number of devices in the IoT network, wireless software-defined networking (W-SDN) is widely agreed to be introduced to facilitate network management, such as launching a new intrusion response scheme (IRS) on massive IoT devices. To guarantee the scalability and security of IoT services on the extended devices in the W-SDN-based network, a distributed RL-based IRS is proposed in this article, called game-theoretic actor–critic-based IRS (GTAC-IRS). GTAC-IRS employs a game-theoretic-based response selection matrix, aiming at reducing training time and facilitating the convergence of the response scheme. GTAC-IRS constructs a well-designed state representation of observed environment status, a low-dimension response matrix, and a simplified response selection policy to lower the complexity of the algorithms. Simulation results reveal that benefiting from local environment observation, GTAC-IRS achieves effective intrusion response without sophisticated feature engineering. Instead of the “warm-start” training adopted in conventional RL-based IRS, the low-dimension response matrix in GTAC-IRS can significantly improve the convergence speed. Thus, GTAC-IRS outperforms other popular IRSs in terms of the response time under the circumstance of the node’s behavior changing or malicious nodes ratio changing.

In this work, we proposed a DDI detection method based on molecular structures using graph convolutional networks and deep sets. We proposed a more discriminative convolutional layer compared to conventional GCN and achieved permutation invariant prediction without losing the capability of capturing complicated interactions.

Learning with curriculum has shown great effectiveness in tasks where the data contains noisy (corrupted) labels, since the curriculum can be used to re-weight or filter out noisy samples via proper design. However, obtaining curriculum from a learner itself without additional supervision or feedback deteriorates the effectiveness due to sample selection bias. Therefore, methods that involve two or more networks have been recently proposed to mitigate such bias. Nevertheless, these studies utilize the collaboration between networks in a way that either emphasizes the disagreement or focuses on the agreement while ignores the other. In this paper, we study the underlying mechanism of how disagreement and agreement between networks can help reduce the noise in gradients and develop a novel framework called Robust Collaborative Learning (RCL) that leverages both disagreement and agreement among networks. We demonstrate the effectiveness of RCL on both synthetic benchmark image data and real-world large-scale bioinformatics data.

The millimeter wave (mmWave) band has been considered as an effective way to meet the increasing traffic demand by sufficient bandwidth and high spatial reuse. The harsh propagation experienced at such high frequencies requires a dense base station deployment, which is not feasible to provide wired backhaul due to the unavailability of fiber drops. To address this issue, the integrated access and backhaul (IAB) architecture is proposed by the third generation partnership project (3GPP). In this paper, we investigate the spectrum and power allocation for IAB mmWave enabled cellular network to maximize the network capacity with considering the data rate requirements. A novel resource allocation scheme based on the sequential convex programming approach (RASCPA) is proposed. The simulation results show that our proposed scheme is superior to other schemes. The proposed scheme can increase the network capacity while satisfying the user data rate requirements.

Despite the fact that deep learning has achieved remarkable success in various domains over the past decade, its application in molecular informatics and drug discovery is still limited. Recent advances in adapting deep architectures to structured data have opened a new paradigm for pharmaceutical research. In this survey, we provide a systematic review on the emerging field of graph convolutional networks and their applications in drug discovery and molecular informatics. Typically we are interested in why and how graph convolution networks can help in drug-related tasks. We elaborate the existing applications through four perspectives: molecular property and activity prediction, interaction prediction, synthesis prediction and de novo drug design. We briefly introduce the theoretical foundations behind graph convolutional networks and illustrate various architectures based on different formulations. Then we summarize the representative applications in drug-related problems. We also discuss the current challenges and future possibilities of applying graph convolutional networks to drug discovery.

As the versatile applications emerge, traffic offloading is an urgent issue to improve the performance for the fifth generation (5G) and beyond networks. We focus on the scenario where a device is enabled to transmit to more than one device simultaneously. The device-to-device (D2D) enabled traffic offloading scheme is studied by employing non-orthogonal multiple access (NOMA) and unlicensed access technologies. Our target is to maximize the capacity of the D2D network by optimizing subchannel assignment and power control while guaranteeing the capacity of NOMA-based cellular links and the WiFi system. The formulated problem is a non-convex mixed integer programming problem, which is hard to solve within a rational time. The problem is decomposed into subchannel assignment and power control subproblems. A matching based licensed subchannel allocation algorithm and an unlicensed subchannel access mechanism are proposed. Furthermore, we propose a centralized power control algorithm and a distributed power control algorithm based on global and local information, respectively. Besides, the unlicensed resource management scheme based on Stackelberg game is proposed to achieve the near-optimal utility of both D2D links and the WiFi system. The simulations illustrate that the proposed scheme can increase the throughput of D2D networks efficiently compared with other works.

This paper presents a Rate-based Cell Range Expansion (CRE) for mmWave massive Multiple Input Multiple Output (MIMO) enabled two-tier heterogeneous networks (Het-Nets) where macro base stations (BSs) operate at sub-6 GHz and small BSs operate at mmWave band. We design a new user association scheme with the consideration of disparities between macro BSs and small BSs, chiefly bandwidths combined with Signal to Interference plus Noise Ratio (SINR), although conventional CRE provides a fixed Cell Selection Offset (CSO) for all user equipments (UEs). A downlink model is put forth to analyze the performance of the proposed Rate-based CRE. Our scheme consists of two parts: CRE between small cells and CRE between macro cells and small cells. We also investigate the proposed method by evaluating the average user throughput and cell edge user throughput. System-level computer simulation results such as average user throughput and 5-percentile user throughput are provided. The results confirm that the proposed method can improve the average user throughput compared with the conventional CRE scheme and maintain the cell edge user throughput for such multi-frequency cooperative networks. Thus, this scheme is an effective solution for user association in 5G mmWave HetNets.

Over the past decade a wide spectrum of machine learning models have been developed to model the neurodegenerative diseases, associating biomarkers, especially non-intrusive neuroimaging markers, with key clinical scores measuring the cognitive status of patients. Multi-task learning (MTL) has been commonly utilized by these studies to address high dimensionality and small cohort size challenges. However, most existing MTL approaches are based on linear models and suffer from two major limitations: 1) they cannot explicitly consider upper/lower bounds in these clinical scores; 2) they lack the capability to capture complicated non-linear interactions among the variables. In this paper, we propose Subspace Network, an efficient deep modeling approach for non-linear multi-task censored regression. Each layer of the subspace network performs a multi-task censored regression to improve upon the predictions from the last layer via sketching a low-dimensional subspace to perform knowledge transfer among learning tasks. Under mild assumptions, for each layer the parametric subspace can be recovered using only one pass of training data. Empirical results demonstrate that the proposed subspace network quickly picks up the correct parameter subspaces, and outperforms state-of-the-arts in predicting neurodegenerative clinical scores using information in brain imaging.

The surging availability of electronic medical records (EHR) leads to increased research interests in medical predictive modeling. Recently many deep learning based predicted models are also developed for EHR data and demonstrated impressive performance. However, a series of recent studies showed that these deep models are not safe: they suffer from certain vulnerabilities. In short, a well-trained deep network can be extremely sensitive to inputs with negligible changes. These inputs are referred to as adversarial examples. In the context of medical informatics, such attacks could alter the result of a high performance deep predictive model by slightly perturbing a patient's medical records. Such instability not only reflects the weakness of deep architectures, more importantly, it offers a guide on detecting susceptible parts on the inputs. In this paper, we propose an efficient and effective framework that learns a time-preferential minimum attack targeting the LSTM model with EHR inputs, and we leverage this attack strategy to screen medical records of patients and identify susceptible events and measurements. The efficient screening procedure can assist decision makers to pay extra attentions to the locations that can cause severe consequence if not measured correctly. We conduct extensive empirical studies on a real-world urgent care cohort and demonstrate the effectiveness of the proposed screening.

Heterogeneous small cell network is a promising technique in the next generation mobile communications. Many works have been studied in small cells, including resource allocation and interference mitigation, but most studies didn't consider the quality-of-service (QoS) and power consumption. This paper focuses on the power allocation based on non-cooperative scheme to mitigate the interference and increase the energy efficiency in small cells. The delay constraint is introduced in small cells to guarantee the QoS. We reconsider the capacity according to Shannon' capacity formula and bring in the concept of effective capacity. We take the total power consumption of the small cells into account and employ energy efficiency metric to formulate the problem of power allocation. The power allocation problem is modeled as non-cooperative supermodular game, and it is shown to converge to Nash equilibrium, and then it is transformed into a convex optimization problem, which is solved by the multi-agent Q-learning algorithm based on conjecture. The effectiveness of the proposed supermodular game based power allocation is verified by the simulations.

Alzheimer's disease (AD), one of the most common causes of dementia, is a severe irreversible neurodegenerative disease that results in loss of mental functions. The transitional stage between the expected cognitive decline of normal aging and AD, mild cognitive impairment (MCI), has been widely regarded as a suitable time for possible therapeutic intervention. The challenging task of MCI detection is therefore of great clinical importance, where the key is to effectively fuse predictive information from multiple heterogeneous data sources collected from the patients. In this paper, we propose a framework to fuse multiple data modalities for predictive modeling using deep matrix factorization, which explores the non-linear interactions among the modalities and exploits such interactions to transfer knowledge and enable high performance prediction. Specifically, the proposed collective deep matrix factorization decomposes all modalities simultaneously to capture non-linear structures of the modalities in a supervised manner, and learns a modality specific component for each modality and a modality invariant component across all modalities. The modality invariant component serves as a compact feature representation of patients that has high predictive power. The modality specific components provide an effective means to explore imaging genetics, yielding insights into how imaging and genotype interact with each other non-linearly in the AD pathology. Extensive empirical studies using various data modalities provided by Alzheimer's Disease Neuroimaging Initiative (ADNI) demonstrate the effectiveness of the proposed method for fusing heterogeneous modalities.

UE-to-Network relay leverages the proximity communication between user equipments (UEs) and allows certain UEs to provide relay assistance for others, which can greatly improve system energy efficiency. In this paper, we consider the scenario where UEs suffering from bad channel condition and low battery level can communicate with the base station directly or via the help of other UEs in heterogeneous networks. The optimal power allocation and connectivity among UEs are studied, which aims at minimizing the system transmission energy while guaranteeing the minimum data rate requirement of each UE. An optimization framework is presented to formulate the system transmission energy minimization problem, which can be converted into a weighted one-to-one matching problem. And a practical joint transmission mode with relay selection and power allocation (JMRP) algorithm is developed to solve it. Simulation results show that the proposed algorithm outperforms the existing works in terms of system transmission energy and throughput.