Health recommender systems (HRSs) enhance prognostication by leveraging clinical information. Existing HRSs often fail to capture the intrinsic correlations between patient phenotypes with similar clinical profiles, necessitating approaches that incorporate patient similarity into prognostic modeling. This work explores such correlations within three classes of biomedical information: diagnosis, procedure, and medication, and the clinical temporality during patient visits to improve diagnostic accuracy. Our approaches include both static and dynamic scenarios. In the static scenario, we propose SIM-PR, which integrates patient similarity and PageRank centrality on a personalized patient graph, and can operate with or without sequential hospital visit information. In the dynamic setting, we develop temporal prediction models based on a multilayer perceptron and a long short-term memory network to learn evolving diagnostic patterns from longitudinal visit histories. Standard supervised machine learning, including logistic regression, random forest, and support vector machines, is employed as comparative baselines. Experiments on the MIMIC-III database demonstrate that the proposed static and temporal models effectively reduce false positives and negatives and achieve superior predictive accuracy compared to existing HRS approaches.

Structural hole spanners (SHSs) are nodes that connect different communities to facilitate efficient information dissemination in complex networks. Existing efforts to identify SHS nodes have predominantly focused on undirected networks, rendering them inadequate to capture directional data flow. This paper presents a novel lightweight approach to motif span scores, called mSpan that leverages network substructures called feed forward loop (FFL) motifs, to detect SHS in directed, weighted as well as unweighted social networks. The proposed approach measures the spanning score of a node in terms of its participation in FFL motifs that bridge network communities. Our theoretical analysis establishes a strong association between the variants of the scores for a given node and the likelihood of its removal disrupting connectivity. We also utilize mSpan to detect spanner motifs that bridge the structural holes in social networks. We validate the efficacy of mSpan in detecting SHS in practical scenarios through comparative evaluations of three real-world social networks against existing spanner detection metrics.

The diversity in disease profiles and therapeutic approaches between hospitals and health professionals underscores the need for patient-centric personalized strategies in healthcare. Alongside this, similarities in disease progression across patients can be utilized to improve prediction models in survival analysis. The need for patient privacy and the utility of prediction models can be simultaneously addressed in the framework of Federated Learning (FL). This paper outlines an approach in the domain of federated survival analysis, specifically the Cox Proportional Hazards (CoxPH) model, with a specific focus on mitigating data heterogeneity and elevating model performance. We present an FL approach that employs feature-based clustering to enhance model accuracy across synthetic datasets and real-world applications, including the Surveillance, Epidemiology, and End Results (SEER) database. Furthermore, we consider an event-based reporting strategy that provides a dynamic approach to model adaptation by responding to local data changes. Our experiments show the efficacy of our approach and discuss future directions for a practical application of FL in healthcare.

Network inference is used to model transcriptional, signaling, and metabolic interactions among genes, proteins, and metabolites that identify biological pathways influencing disease pathogenesis. Advances in machine learning (ML)-based inference models exhibit the predictive capabilities of capturing latent patterns in genomic data. Such models are emerging as an alternative to the statistical models identifying causative factors driving complex diseases. We present CoVar, an ML-based framework that builds upon the properties of existing inference models, to find the central genes driving perturbed gene expression across biological states. Unlike differentially expressed genes (DEGs) that capture changes in individual gene expression across conditions, CoVar focuses on identifying variational genes that undergo changes in their expression network interaction profiles, providing insights into changes in the regulatory dynamics, such as in disease pathogenesis. Subsequently, it finds core genes from among the nearest neighbors of these variational genes, which are central to the variational activity and influence the coordinated regulatory processes underlying the observed changes in gene expression. Through the analysis of simulated as well as yeast expression data perturbed by the deletion of the mitochondrial genome, we show that CoVar captures the intrinsic variationality and modularity in the expression data, identifying key driver genes not found through existing differential analysis methodologies.

Vaccines have proven useful in curbing contagion from new strains of the SARS-CoV-2 virus. However, equitable vaccine allocation continues to be a significant challenge worldwide, necessitating a comprehensive allocation strategy incorporating heterogeneity in epidemiological and behavioral considerations. In this paper, we present a hierarchical allocation strategy that assigns vaccines to zones and their constituent neighborhoods cost-effectively, based on their population density, susceptibility, infected count, and attitude towards vaccinations. Moreover, it includes a module that tackles vaccine shortages in certain zones by locally transferring vaccines from zones with surplus vaccines. We leverage the epidemiological, socio-demographic, and social media datasets from Chicago and Greece and their constituent community areas to show that the proposed allocation approach assigns vaccines based on the chosen criteria and captures the effects of disparate vaccine adoption rates. We conclude the paper with a lowdown on future efforts to extend this study to design models for effective public policies and vaccination strategies that curtail vaccine purchase costs.

COVID-19 is a global pandemic caused by the Severe Acute Respiratory Syndrome Coronavirus 2. While swift vaccine development and distribution have arrested the infection spread rate, it is necessary to design public policies that inform human mobility to curb outbreaks from future strains of the virus. While existing non-pharmaceutical approaches employing network science and machine learning offer promising travel policy solutions, they are guided by epidemiological and economic considerations alone and not human itineraries. We introduce an evolutionary algorithm (EA) based mobility scheduler that incorporates the personalized itineraries of individuals to determine the ideal timing of their mobility. We mathematically analyze the computational efficiency versus the optimality trade-off of the mobility scheduler. Through extensive simulations, we demonstrate that the EA-based mobility scheduler can balance the trade-off between (1) optimality and computational cost and (2) fair and preferential human mobility while reducing contagion under lockdown and no-lockdown as well as even and uneven human mobility traffic scenarios. We show that for two human mobility models, the scheduler exhibits lower infection numbers than a baseline trip-planning approach that directs human traffic along the least congested route to minimize contagion. We discuss that the EA scheduler lends itself to intricate mobility schedules of multiple destination choices with varying priorities and socioeconomic and demographic considerations.

Rapid development and distribution of vaccines have been a hallmark of the battle against COVID-19. While the efficacy, clinical trials, adverse health effects, and sociodemographic and clinical factors determining the distribution of vaccines have been studied extensively, there has been little effort to design cost-effective vaccine provisioning schemes. We introduce a vaccine provisioning scheme that leverages coalitional game theory to improve the cost of vaccines while meeting the epidemiological demand of neighboring zones. The proposed approach incentivizes bulk purchases by groups (or coalitions) of zones at lower prices while penalizing large coalitions to avoid logistical challenges. Moreover, it enables the policymaker to model the vaccine demand of zones based on their epidemiological profiles, such as susceptible, infected numbers or population density, or a combination thereof. We carry out experiments using the SEIRD (susceptible, exposed, infected, recovered, death) epidemic model as well as the daily confirmed cases in the five boroughs of New York City to show the efficacy of the approach.

Internet of Things (IoT) enables the collection of large volumes of data by billions of pervasive intelligent devices and sharing them with remote cloud servers for processing, resulting in increasing network congestion and server response times. The advent of edge computing has addressed these challenges by introducing an intermediate edge layer comprising networked fog nodes that provide on-demand computation, caching, and communication services to meet critical Quality-of-Service (QoS) requirements. However, in a challenging environment brought by disaster and aftershocks, the QoS is hampered as several fog nodes in the edge layer are damaged. Nevertheless, the existing network infrastructure should still support the uninterrupted flow of time-critical contextual data between the survivors and rescuers for quick recovery operations. In this work, we envision that the IoT devices and existing fog nodes will collaborate to form ad-hoc networks for emergency message delivery under disaster situations. We present a distributed routing mechanism, termed motif centrality-based routing ($MCR$), that leverages the concept of network motifs (subgraphs) seen in social and biological networks. Specifically, the proposed mechanism addresses three QoS requirements of an ad-hoc network: 1) robustness against component failures; 2) low latency; and 3) energy efficiency. We experimentally show that the$MCR$-based routing ensures high data delivery, low latency, and comparable efficiency in energy usage. Finally, an extensive simulation-based study shows that$MCR$outperforms the related benchmarks in terms of the three QoS requirements.

Unmanned Aerial Vehicles (UAVs) can be utilized as aerial base stations to establish wireless communication networks in various challenging scenarios, such as emergency disaster areas and rural areas. Under large regions, the aerial communication networks would require UAVs to form wireless (backhaul) links among each other to provide end-to-end wireless services between two or more ground users (via one or more UAVs). Such UAV backhauling in aerial communication networks may be severely compromised if one or more UAVs are knocked off during the time of operation – it may be due to UAV hardware/software faults, limited battery, malicious attacks, etc. Deep reinforcement learning (DRL) has emerged as a powerful tool for learning tasks with large state and continuous action spaces. In this paper, we leverage emerging DRL to achieve reliable backhauling in an aerial communication network that remains functional and supports end-to-end wireless services even under various random and/or targeted UAV node failures. The proposed method (i) maximizes the reliability of UAV backhauling with joint consideration for communication coverage, (ii) learns the complex environment and its dynamics, and (iii) makes 3D positioning decisions for each UAV under the guidance of two deep neural networks. Our performance evaluation reveals that the proposed DRL approach outperforms the baseline method in terms of wireless coverage and network reliability against UAV failures.

Vaccines have proved to be highly effective in preventing severe outcomes in COVID-19 patients. Despite swift vaccine development, the policymakers are still struggling to meet the global challenges in the availability, cost and distribution of vaccines. With the emergence of new vaccine types and boosters to beat the newer strains of the virus, it is necessary to design effective vaccine distribution strategies. In this paper, we present generalizable, multi-vaccine distribution measures that allocate vaccines based on the socio-economic, epidemiological and demographic profiles of different zones. The proposed approach incorporates myriad features, whereby it can assign vaccines based on a subset of the chosen criteria, minimize or fix the number of assigned vaccines and balance the trade-off between cost and criteria. Through simulation experiments, we demonstrate the ability of the optimizer to capture the variable vaccine adoption rates among zones and reward lower vaccine hesitancy with reduced contagion.

Wireless Sensor Networks (WSNs) enable promising solutions to large-scale industrial, medical and environmental tracking and monitoring applications. Software Defined Networking (SDN) is a new paradigm that decouples the network control and data forwarding functionalities and may potentially improve data sensing in a highly dynamic environment. The networking community is directing its efforts towards ensuring that the software defined WSNs (SD-WSN) carry out the data sensing even in a hostile environment characterized by node or link failures. In this work, we present an adaptive topology control strategy based on reinforcement learning for mobile SD-WSN. The approach employs the notion of statistically significant subgraphs, called motifs, that have been shown to render graph robustness to biological networks. Our simulation experiments on the map of New York City shows that this approach is capable of modulating network parameters to achieve varying goals such as high data delivery, low latency and energy efficiency.

Software Defined Wireless Sensor Networks (SD-WSN) is a promising paradigm in wireless communication that offers high flexibility in network management by enabling dynamic and programmable network control. SDN controller has a centralized global view of the network, making it an ideal choice for data sensing in a highly dynamic sensing environment. We proposed a reinforcement learning (RL) based adaptive topology control approach (Roy et al., IEEE CCNC 2020) that employs periodic node mobility to meet diverse network objectives, such as data delivery, latency, and energy efficiency. We also demonstrated that erratic mobility can considerably hamper the learning of the RL module resulting in poor overall quality of service. In this work, we present a customized network simulation environment that captures the variations in the performance of the proposed SD-WSN framework. Finally, we present a new approach based on supervised machine learning that can identify periodic mobility and mitigate the ill-effects of erratic mobility.

Selective removal of certain subgraphs called motifs based on the spread function value is one of the most powerful approaches to curb the overall influence spread in any complex network. In this paper, we first prove that any general spread function preserves both monotonicity and submodularity properties even under motif removal operations. Next, we propose a scoring mechanism as a novel spread function that quantifies the relative importance of a given motif within the overall influence spread dynamics on the complex network. We design a novel algorithm that eliminates motifs with high spread scores to curb influence spread. We evaluate the performance of our proposed spread control algorithm using simulation experiments in the context of 3-node motifs called feed forward loops (FFLs) in both real and synthetic network topologies. We demonstrate that high-scoring motifs intercept a high number of short paths from the pre-assigned source and sinks, because of which their elimination results in a significant effect on curbing the influence spread. Furthermore, we empirically evaluate the run-time and cost versus performance trade-off of the proposed algorithm.

Unmanned Aerial Vehicles (UAVs), as aerial base stations, is a promising solution for providing end-to-end wireless communications to ground users, thanks to its positioning, flexibility, and autonomy. However, to provide end-to-end wireless communication services, all UAVs must ensure a reliable multi-UAV network topology, even when one or more UAVs are knocked off the network due to hardware/software faults, unreliable wireless connections, etc. Hence, how to design a reliable Multi-UAV network with a minimum number of UAVs becomes a key design issue, which is largely unaddressed in the literature. In this paper, we propose exploring biological robustness to design a reliable MuLtI-UAV NetworK, termed, bio-LINK, which is resilient against the UAV node failures and thus, ensures reliable end-to-end communication services to ground users. We first formulate the above bio-LINK problem as an integer linear programming (ILP) optimization problem and show it is NP-Hard. Next, we propose a polynomial-time heuristic that employs an iterative UAV positioning inspired by Markov Chain Monte Carlo (MCMC) random sampling approach. Our extensive simulation study shows that the proposed algorithm outperforms three baseline algorithms in terms of several considered robustness metrics (e.g., motif count, network efficiency, etc.) and ground user coverage, notwithstanding the random and targeted failure of UAV nodes. When compared with a baseline algorithm with the same number of UAVs, the proposed algorithm retains the motif count by 5-6 folds and improves network efficiency by 39 - 95% and ground user coverage by 2-18%.

In the aftermath of large-scale disasters, such as earthquakes or hurricanes, existing communication infrastructures are often critically impaired, preventing timely information exchange between the survivors, responders, and the coordination center. Smart devices, movable base stations, easily deployable WiFi routers, and unimpaired communication towers can be used to set up temporary networks, called disaster response networks (DRNs). However, such networks are challenged by rapid energy depletion of smart devices as well as component failures. To address these issues, in this paper we propose a novel energy-efficient yet robust DRN topology, termed Bio-DRN, that mimics the inherent robustness of a biological network of living organisms, called gene regulatory network (GRN). Specifically, the Bio-DRN is a subgraph of the DRN topology generated by one-to-one mapping between the structurally similar genes and DRN components, i.e., survivors, points of interest like shelter points, and the coordination center. We first formulate the construction of Bio-DRN topology as an integer linear programming optimization problem, and show that it is NP-hard. Then, we present a sub-optimal heuristic that constructs the Bio-DRN topology as a common subgraph of both GRN and DRN topologies. Our experimental study on a real disaster prone region in Bhaktapur, Nepal, shows that Bio-DRN preserves the topological properties of GRN, such as low graph density and motif abundance, and achieves both energy efficiency and network robustness, while ensuring timely message delivery.

Recent years have seen a proliferation of intelligent (automated) decision support systems for various smart city applications such as energy management, transportation, healthcare, environment monitoring, and so on. A key enabler in the smart city paradigm is the Internet-of-Things (IoT) network of smart sensing and actuation devices assisting in real-time detection and monitoring of physical phenomena. The underlying IoT network must be energy-efficient for application sustainability and also quality of information (QoI)-aware for near-perfect device actuation. To this end, this paper proposes bioSmartSense, a novel bio-inspired distributed event sensing and data collection framework, based on the gene regulatory networks (GRNs) in living organisms. The idea is to make the sensing and reporting tasks energy-efficient through self-modulation of IoT device energy levels, analogous to the activation or repression of genes by the regulating proteins, called Transcription Factors (TFs). To support energy-efficient and QoI-aware information dissemination, we first customize a heuristic designed for the Maximum Weighted Independent Set problem encompassing both `quality' and `quantity' of sensed data, where the former depends on the device energy levels while the latter on the number of events sensed. We utilize the heuristic to propose a sub-optimal device selection mechanism constrained on the IoT network's overall residual energy. Simulation experiments demonstrate that the bioSmartSense framework achieves better energy-efficiency while maximizing event reporting compared to a state-of-the-art data collection approach for smart city applications.

Assessing risk for voluminous legal documents such as request for proposal, contracts is tedious and error prone. We have developed "risk-o-meter", a framework, based on machine learning and natural language processing to review and assess risks of any legal document. Our framework uses Paragraph Vector, an unsupervised model to generate vector representation of text. This enables the framework to learn contextual relations of legal terms and generate sensible context aware embedding. The framework then feeds the vector space into a supervised classification algorithm to predict whether a paragraph belongs to a pre-defined risk category or not. The framework thus extracts risk prone paragraphs. This technique efficiently overcomes the limitations of keyword based search. We have achieved an accuracy of 91% for the risk category having the largest training dataset. This framework will help organizations optimize effort to identify risk from large document base with minimal human intervention and thus will help to have risk mitigated sustainable growth. Its machine learning capability makes it scalable to uncover relevant information from any type of document apart from legal documents, provided the library is pre-populated and rich.

Gene Regulatory Networks (GRNs) are biological networks that have been widely studied for their ability to regulate protein synthesis in cells by robust signal propagation. The innate biological robustness of GRN is attributed to the occurrence of statistically significant subgraphs, called motifs. While Wireless Sensor Network (WSN) topologies designed using GRN graphs, called bio-WSNs, have been proven to exhibit significant improvement in packet delivery and network latency over random graph-based WSNs, it is still not clear what role motifs play in the observed performance improvement of bio-WSNs. This work explores why a dominant 3-node motif, called Feed Forward Loop (FFL), typifies the robustness of GRN motifs. We also employ graph centrality metrics to corroborate biological studies that have shown motifs to provide pathways for signal propagation in GRNs. Finally, we perform graph-theoretic and simulation experiments on GRN subgraphs and their corresponding bio-WSNs to demonstrate that nodes with high FFL motif participation offer multiple short and robust communication pathways, despite the failure of random and targeted nodes and links.

Large scale disasters require prompt rescue and relief operations to restrict further casualties. To carry out such operations, it is essential to have a communication infrastructure between survivors and responders, which is often impaired due to the disaster. Off-the-shelf wireless devices such as smartphones, PDAs and Laptops offer an effective solution towards the establishment of makeshift communication infrastructure. However, in the absence of bonafide power sources, it becomes imperative to judiciously utilize energy (battery power) of such devices such that the network is functional until primary infrastructure is restored. This paper proposes a novel approach, called Cluster based Topological Routing (CTR) that prolongs the longevity of the network by exploiting the natural gathering of survivors in shelter points. In particular, the clustering algorithm identifies such survivor groups combined with a data forwarding approach, to minimize the number of data transmissions yet guaranteeing the required packet delivery and network latency. Our extensive simulation study shows that CTR yields twice the network lifetime than existing routing approaches in disaster response networks, while ensuring comparable packet delivery and network latency.