Propagation Network Research Articles

BiFSMNv2: Pushing Binary Neural Networks for Keyword Spotting to Real-Network Performance.

Deep neural networks, such as the deep-FSMN, have been widely studied for keyword spotting (KWS) applications while suffering expensive computation and storage. Therefore, network compression technologies such as binarization are studied to deploy KWS models on edge. In this article, we present a strong yet efficient binary neural network for KWS, namely, BiFSMNv2, pushing it to the real-network accuracy performance. First, we present a dual-scale thinnable 1-bit-architecture (DTA) to recover the representation capability of the binarized computation units by dual-scale activation binarization and liberate the speedup potential from an overall architecture perspective. Second, we also construct a frequency-independent distillation (FID) scheme for KWS binarization-aware training, which distills the high- and low-frequency components independently to mitigate the information mismatch between full-precision and binarized representations. Moreover, we propose the learning propagation binarizer (LPB), a general and efficient binarizer that enables the forward and backward propagation of binary KWS networks to be continuously improved through learning. We implement and deploy BiFSMNv2 on ARMv8 real-world hardware with a novel fast bitwise computation kernel (FBCK), which is proposed to fully use registers and increase instruction throughput. Comprehensive experiments show our BiFSMNv2 outperforms the existing binary networks for KWS by convincing margins across different datasets and achieves comparable accuracy with the full-precision networks (only a tiny 1.51% drop on Speech Commands V1-12). We highlight that benefiting from the compact architecture and optimized hardware kernel, BiFSMNv2 can achieve an impressive 25.1× speedup and 20.2× storage-saving on edge hardware.

Channel assisted noise propagation in a two-step cascade.

Signal propagation in biochemical networks is characterized by the inherent randomness in gene expression and fluctuations of the environmental components, commonly known as intrinsic and extrinsic noise, respectively. We present a theoretical framework for noise propagation in a generic two-step cascade (S→X→Y) regarding intrinsic and extrinsic noise. We identify different channels of noise transmission that regulate the individual and the overall noise properties of each component. Our analysis shows that the intrinsic noise of S alleviates the general noise and information transmission capacity along the cascade. On the other hand, the intrinsic noise of X and Y acts as a bottleneck of information transmission. We also show a hierarchical relationship among the intrinsic noise levels of S, X, and Y, with S exhibiting the highest level of intrinsic noise, followed by X and then Y. This hierarchy is preserved within the two-step cascade, facilitating the highest information transmission from S to Y via X.

Coplane-constrained sparse depth sampling and local depth propagation for depth estimation

Depth estimation with sparse reference has emerged recently, and predicts depth map from a monocular image and a set of depth reference samples. Previous works randomly select reference samples by sensors, leading to severe depth bias as this sampling is independent to image semantic and neglects the unbalance of depth distribution in regions. This paper proposes a Coplane-Constrained sparse Depth (CCD) sampling to explore representative reference samples, and design a Local Depth Propagation (LDP) network for complete the sparse point cloud map. This can capture diverse depth information and diffuse the valid points to neighbors with geometry prior. Specifically, we first construct the surface normal map and detect coplane pixels by superpixel segmenting for sampling references, whose depth can be represented by that of superpixel centroid. Then, we introduce local depth propagation to obtain coarse-level depth map with geometric information, which dynamically diffuses the depth from the reference to neighbors based on local planar assumption. Further, we generate the fine-level depth map by devising a pixel-wise focal loss, which imposes the semantic and geometry calibration on pixels with low confidence in coarse-level prediction. Extensive experiments on public datasets demonstrate that our model outperforms SOTA depth estimation and completion methods.

Analyzing risk influencing factors of ship collision accidents: A data-driven Bayesian network model integrating physical knowledge

The interaction and propagation effects among risk-influencing factors (RIFs) often lead to maritime accidents. Therefore, to support maritime risk management in decision-making, it is imperative to conduct a quantitative risk assessment (QRA) of these accidents, which requires a methodology capable of capturing causal relationships in limited cases. To achieve this goal, this study proposes a data-driven Bayesian network (BN) model that integrates physical knowledge with the QRA. Based on the collected data, a combination of domain knowledge, structure learning, and parameter learning is employed to construct the model. In the data-driven phase, three structural learning algorithms including Bayesian search (BS), Greedy thick thinning (GTT), and Peter-Clark (PC) algorithms, were used in combination with the expectation maximization algorithm of parameter learning. Then, using four indices calculated by the confusion matrix, the most fitting algorithm for structure learning was chosen, and the confusion matrix was obtained by five-fold cross-validation. Finally, network propagation impact (NPI) is introduced to prioritize RIFs. Preventive measures and emergency plans are formulated based on the network resilience metric (NRM) and network vulnerability metric (NVM). This approach leverages the objectivity of a data-driven BN and integrates domain expertise to enhance model validity. A case study of collisions was conducted to demonstrate the applicability of the model. The data were sourced from 327 collision accident reports provided by the China Maritime Safety Administration, and the PC algorithm was chosen. Findings indicate that the “Management system”, “Supervision”, and “Crew training” RIFs are prioritized for preventive measures, whereas “Communication” receives more resources for emergency handling. The results suggest the formulation of risk management strategies for practitioners to improve maritime safety.

A network-based trans-omics approach for predicting synergistic drug combinations

BackgroundCombination therapy can offer greater efficacy on medical treatments. However, the discovery of synergistic drug combinations is challenging. We propose a novel computational method, SyndrumNET, to predict synergistic drug combinations by network propagation with trans-omics analyses.MethodsThe prediction is based on the topological relationship, network-based proximity, and transcriptional correlation between diseases and drugs. SyndrumNET was applied to analyzing six diseases including asthma, diabetes, hypertension, colorectal cancer, acute myeloid leukemia (AML), and chronic myeloid leukemia (CML).ResultsHere we show that SyndrumNET outperforms the previous methods in terms of high accuracy. We perform in vitro cell survival assays to validate our prediction for CML. Of the top 17 predicted drug pairs, 14 drug pairs successfully exhibits synergistic anticancer effects. Our mode-of-action analysis also reveals that the drug synergy of the top predicted combination of capsaicin and mitoxantrone is due to the complementary regulation of 12 pathways, including the Rap1 signaling pathway.ConclusionsThe proposed method is expected to be useful for discovering synergistic drug combinations for various complex diseases.

Solute transport characteristics of columnar volumetric contraction networks with mega column structure and aperture variability

Numerical simulations explore for the first time the role of mega columns and aperture variability on particle transport through mature volumetric contraction networks as informed by a unique synthesis of network propagation and maturity. Columnar fracture patterns are generated by updating a series of Voronoi centers to the midpoint of a generated polygon over many iterations, creating 250 network realizations. A DFN simulator solves for fluid flow and tracks conservative particle trajectories within each network. Dominant fracture attributes impacting fluid flow and solute transport in volumetric contraction networks are fracture orientation, density, and aperture/transmissivity. Ensemble plume snapshots generated by networks with equal fracture transmissivity define a baseline-level of dispersion that is solely attributed to network structure and connectivity. Longitudinal and transverse dispersion increase and the center of plume mass becomes delayed relative to the baseline case when fracture transmissivity is varied according to a lognormal distribution. The incorporation of highly-transmissive, large-aperture mega column fractures leads to plume snapshots with a more pronounced leading edge and an order of magnitude faster average breakthrough times. The breakthrough curves contain three peaks reflecting contrasting transport pathways in which particles are: (i) initially placed in mega column fractures and remain in these features until exiting the model domain, (ii) initially placed into small column fractures, incur additional time to migrate and enter a mega column fracture, and remain within those mega columns, and (iii) initially placed in small column fractures and remain in these fractures. Incorporating variability in fracture transmissivity for both small column and mega column fractures disrupts the binary distinction between small column and mega column fracture velocities and leads to dispersed breakthroughs over long time scales with a single peak. These results demonstrate that preferential flow paths emerge in volumetric contraction networks due to contracts in fracture transmissivity, not fracture connectivity as observed in tectonic networks.

Research on the Influence of Information Iterative Propagation on Complex Network Structure.

Dynamic propagation will affect the change of network structure. Different networks are affected by the iterative propagation of information to different degrees. The iterative propagation of information in the network changes the connection strength of the chain edge between nodes. Most studies on temporal networks build networks based on time characteristics, and the iterative propagation of information in the network can also reflect the time characteristics of network evolution. The change of network structure is a macromanifestation of time characteristics, whereas the dynamics in the network is a micromanifestation of time characteristics. How to concretely visualize the change of network structure influenced by the characteristics of propagation dynamics has become the focus of this article. The appearance of chain edge is the micro change of network structure, and the division of community is the macro change of network structure. Based on this, the node participation is proposed to quantify the influence of different users on the information propagation in the network, and it is simulated in different types of networks. By analyzing the iterative propagation of information, the weighted network of different networks based on the iterative propagation of information is constructed. Finally, the chain edge and community division in the network are analyzed to achieve the purpose of quantifying the influence of network propagation on complex network structure.

Comparative network-based analysis of toll-like receptor agonist, L-pampo signaling pathways in immune and cancer cells

Toll-like receptors (TLRs) are critical components to stimulate immune responses against various infections. Recently, TLR agonists have emerged as a promising way to activate anti-tumor immunity. L-pampo, a TLR1/2 and TLR3 agonist, induces humoral and cellular immune responses and also causes cancer cell death. In this study, we investigated the L-pampo-induced signals and delineated their interactions with molecular signaling pathways using RNA-seq in immune cells and colon and prostate cancer cells. We first constructed a template network with differentially expressed genes and influential genes from network propagation using the weighted gene co-expression network analysis. Next, we obtained perturbed modules using the above method and extracted core submodules from them by conducting Walktrap. Finally, we reconstructed the subnetworks of major molecular signals utilizing a shortest path-finding algorithm, TOPAS. Our analysis suggests that TLR signaling activated by L-pampo is transmitted to oxidative phosphorylation (OXPHOS) with reactive oxygen species (ROS) through PI3K-AKT and JAK-STAT only in immune and prostate cancer cells that highly express TLRs. This signal flow may further sensitize prostate cancer to L-pampo due to its high basal expression level of OXPHOS and ROS. Our computational approaches can be applied for inferring underlying molecular mechanisms from complex gene expression profiles.

Unveiling hidden connections in omics data via pyPARAGON: an integrative hybrid approach for disease network construction.

Network inference or reconstruction algorithms play an integral role in successfully analyzing and identifying causal relationships between omics hits for detecting dysregulated and altered signaling components in various contexts, encompassing disease states and drug perturbations. However, accurate representation of signaling networks and identification of context-specific interactions within sparse omics datasets in complex interactomes pose significant challenges in integrative approaches. To address these challenges, we present pyPARAGON (PAgeRAnk-flux on Graphlet-guided network for multi-Omic data integratioN), a novel tool that combines network propagation with graphlets. pyPARAGON enhances accuracy and minimizes the inclusion of nonspecific interactions in signaling networks by utilizing network rather than relying on pairwise connections among proteins. Through comprehensive evaluations on benchmark signaling pathways, we demonstrate that pyPARAGON outperforms state-of-the-art approaches in node propagation and edge inference. Furthermore, pyPARAGON exhibits promising performance in discovering cancer driver networks. Notably, we demonstrate its utility in network-based stratification of patient tumors by integrating phosphoproteomic data from 105 breast cancer tumors with the interactome and demonstrating tumor-specific signaling pathways. Overall, pyPARAGON is a novel tool for analyzing and integrating multi-omic data in the context of signaling networks. pyPARAGON is available at https://github.com/netlab-ku/pyPARAGON.

An Early Warning Model for Turbine Intermediate-Stage Flux Failure Based on an Improved HEOA Algorithm Optimizing DMSE-GRU Model

As renewable energy sources such as wind and photovoltaics continue to enter the grid, their intermittency and instability leads to an increasing demand for peaking and frequency regulation. An efficient dynamic monitoring method is necessary to improve the safety level of intelligent operation and maintenance of power stations. To overcome the insufficient detection accuracy and poor adaptability of traditional methods, a novel fault early warning method with careful consideration of dynamic characteristics and model optimization is proposed. A combined loss function is proposed based on the dynamic time warping and the mean square error from the perspective of both shape similarity and time similarity. A prediction model of steam turbine intermediate-stage extraction temperature based on the gate recurrent unit is then proposed, and the change in prediction residuals is utilized as a fault warning criterion. In order to further improve the diagnostic accuracy, a human evolutionary optimization algorithm with lens opposition-based learning is proposed for model parameter adaptive optimization. Experiments on real-world normal and faulty operational data demonstrate that the proposed method can improve the detection accuracy by an average of 1.31% and 1.03% compared to the long short-term memory network, convolutional neural network, back propagation network, extreme learning machines, gradient boosting decision tree, and LightGBM models.

Research on the dynamic spread of information in social networks based on relationship strength theory and feedback mechanism

Introduction: Studying the main factors and related paths of rumor propagation contributes to the precise governance of rumor information in social networks. Most existing network representation learning methods do not fit with real-world information propagation networks well, and the network cannot effectively model the temporal characteristics and dynamic evolution features of rumor information propagation.Methods: Our study proposes a new dynamic network representation model for information propagation. Additionally, the study introduces a feedback mechanism where the latest node representations are fed back to neighboring nodes.Results: The method solves the problem of delayed network representation and improves network representation performance.Discussion: We conducted experimental simulations, and the results indicate that a higher level of trust contributes to stable group relationships and converging opinions, reducing the likelihood of opinion dispersion. Furthermore, novelty of topics, and interactivity between users, and opinion leaders exhibit distinct characteristics in guiding public opinion. The viewpoint evolution of the newly constructed dynamic network representation model aligns more closely with viewpoint evolution in real-world social networks.

APPN: An Attention-based Pseudo-label Propagation Network for few-shot learning with noisy labels

Few-shot learning has garnered significant attention in deep learning as an effective approach for addressing the issue of data scarcity. Conventionally, training datasets in few-shot learning are clean. However, due to the occurrence of sensor malfunctions, data transmission anomalies, or inaccuracies in manual annotation, the accuracy of sample annotations cannot be guaranteed. Therefore, dealing with few-shot learning with noisy label becomes an urgent problem. To solve the above problem, we propose an Attention-based Pseudo-label Propagation Network (APPN). We make some technical contributions: (1) We propose an attention-based feature extraction method that can effectively capture the distinctions between clean and noisy samples. (2) We propose an improved graph-based pseudo-label propagation method that utilizes pseudo-labels with latent class information as initial labels, thereby enhancing the accuracy of label propagation. (3) We describe a comprehensive multi-step noise detection method which can accurately detect noisy samples from clean samples and effectively distinguish between out-of-domain (OOD) noise and in-domain (ID) noise. (4) Finally, the solid experiments verify the better performance on FC100, CIFAR-FS, miniImageNet, and tieredImageNet datasets. The results show that our method has high robustness and superiority in the few-shot learning with noisy labels. Code and models at: https://github.com/Typistchen/APPN.

Risk amplification effect of multilayer financial networks: Feedback mechanism or cyclic structure?

This paper analyzes the amplification characteristics of risk propagation in multi-layer financial networks from the perspective of network topologies. The research finds that, even without feedback pathways, cross-layer links alone can lead to more severe risk contagion than in single-layer networks. More importantly, with increasing connectivity, the formation of cyclic structures in multi-layer networks will significantly exacerbate the systemic risk.

An In Silico Modelling Approach to Predict Hemodynamic Outcomes in Diabetic and Hypertensive Kidney Disease

Early diagnosis of kidney disease remains an unmet clinical challenge, preventing timely and effective intervention. Diabetes and hypertension are two main causes of kidney disease, can often appear together, and can only be distinguished by invasive biopsy. In this study, we developed a modelling approach to simulate blood velocity, volumetric flow rate, and pressure wave propagation in arterial networks of ageing, diabetic, and hypertensive virtual populations. The model was validated by comparing our predictions for pressure, volumetric flow rate and waveform-derived indexes with in vivo data on ageing populations from the literature. The model simulated the effects of kidney disease, and was calibrated to align quantitatively with in vivo data on diabetic and hypertensive nephropathy from the literature. Our study identified some potential biomarkers extracted from renal blood flow rate and flow pulsatility. For typical patient age groups, resistive index values were 0.69 (SD 0.05) and 0.74 (SD 0.02) in the early and severe stages of diabetic nephropathy, respectively. Similar trends were observed in the same stages of hypertensive nephropathy, with a range from 0.65 (SD 0.07) to 0.73 (SD 0.05), respectively. Mean renal blood flow rate through a single diseased kidney ranged from 329 (SD 40, early) to 317 (SD 38, severe) ml/min in diabetic nephropathy and 443 (SD 54, early) to 388 (SD 47, severe) ml/min in hypertensive nephropathy, showing potential as a biomarker for early diagnosis of kidney disease. This modelling approach demonstrated its potential application in informing biomarker identification and facilitating the setup of clinical trials.

Wave propagation in a light-temperature neural network under adaptive local energy balance.

External electric and mechanical stimuli can induce shape deformation in excitable media because of its intrinsic flexible property. When the signals propagation in the media is described by a neural network, creation of heterogeneity or defect is considered as the effect of shape deformation due to accumulation or release of energy in the media. In this paper, a temperature-light sensitive neuron model is developed from a nonlinear circuit composed of a phototube and a thermistor, and the physical energy is kept in capacitive and inductive terms. Furthermore, the Hamilton energy for this function neuron is obtained in theoretical way. A regular neural network is built on a square array by activating electric synapse between adjacent neurons, and a few of neurons in local area is excited by noisy disturbance, which induces local energy diversity, and continuous coupling enables energy propagation and diffusion. Initially, the Hamilton energy function for a temperature-light sensitive neuron can be obtained. Then, the finite neurons are applied noise to obtain energy diversity to explore the energy spread between neurons in the network. For keeping local energy balance, one intrinsic parameter is regulated adaptively until energy diversity in this local area is decreased greatly. Regular pattern formation indicates that local energy balance creates heterogeneity or defects and a few of neurons show continuous parameter shift for keeping energy balance in a local area, which supports gradient energy distribution for propagating waves in the network.

Investigating the spatial propagation patterns of meteorological drought events and underlying mechanisms using complex network theory: A case study of the Yangtze River Basin, China

The spatial propagation patterns of meteorological drought events (MDEs) and underlying mechanisms contribute to elucidating and forecasting drought evolution. In this study, gridded MDEs in the Yangtze River Basin (YRB) throughout the entire year, wet season and dry season were extracted from 3-month Standardized Precipitation Evapotranspiration Index (SPEI-3) series. Event synchronization (ES) and complex networks (CN) were employed to construct the MDE synchronization networks and MDE spatial propagation networks for various periods. The former were utilized to identify MDE synchronized subregions where MDEs co-occur and co-evolve in the YRB, while the latter were used to quantify the MDE spatial propagation patterns over both the basin and its subregions. The driving mechanisms behind MDE spatial propagation were further investigated by diagnosing the concomitant drought-inducing climate systems. The findings reveal the presence of four MDE synchronized subregions during the wet season and five subregions during the entire year and dry season. These subregions exhibited distinct spatial propagation patterns of MDEs, aligning with overall findings across the YRB. Notable differences were observed between wet and dry seasons, with various subregions exhibiting distinctive spatial propagation patterns during each season. These patterns are driven by variations in the controlling atmospheric circulation systems, leading to anomalies of wind patterns and moisture distribution, ultimately resulting in deficient moisture supply. The variations of tropical sea surface thermal conditions, influences of the Tibetan Plateau and MDE self-propagation triggered by land–atmosphere feedback are considered as three primary influencing factors for MDE spatial propagation in the YRB.

Effects of packetization on communication dynamics in brain networks.

Computational studies in network neuroscience build models of communication dynamics in the connectome that help us understand the structure-function relationships of the brain. In these models, the dynamics of cortical signal transmission in brain networks are approximated with simple propagation strategies such as random walks and shortest path routing. Furthermore, the signal transmission dynamics in brain networks can be associated with the switching architectures of engineered communication systems (e.g., message switching and packet switching). However, it has been unclear how propagation strategies and switching architectures are related in models of brain network communication. Here, we investigate the effects of the difference between packet switching and message switching (i.e., whether signals are packetized or not) on the transmission completion time of propagation strategies when simulating signal propagation in mammalian brain networks. The results show that packetization in the connectome with hubs increases the time of the random walk strategy and does not change that of the shortest path strategy, but decreases that of more plausible strategies for brain networks that balance between communication speed and information requirements. This finding suggests an advantage of packet-switched communication in the connectome and provides new insights into modeling the communication dynamics in brain networks.

Mining Delay Propagation Causality within an Airport Network from Historical Data

Airport networks are interconnected through flight routes, with delays at upstream airports leading to delays at downstream airports, thus causing delay propagation. Exploring the mechanisms of delay propagation in airport networks provides scientific insights for managing and controlling delays in aviation systems. Existing methods, such as Granger causality tests and transfer entropy, must be revised to address the nonlinear causal relationships of delays in airport networks. So, this paper proposes a causality mining method for delay propagation in airport networks based on partial correlation-based multivariate conditional independence (PCMCI). This method comprehensively considers all airports and causality mining in two stages. The first stage uses conditional independence tests to obtain the parent node set of the target airport, which includes both true and false causal relationships. The second stage employs instantaneous conditional independence tests to eliminate false causal relationships and obtain test statistics representing the strength of causality. Based on historical delay data from US airports over a year, the experimental results show that multiple factors cause delay propagation in airport networks rather than a single causal relationship. The scope of delay propagation is limited, mainly affecting a few airports closely connected to it. Delays at airports with small flight volumes are more likely to propagate. Few airport pairs in the network mutually propagate delays and, often, delays at airports affected by a particular airport’s delay also exhibit causal relationships with each other. This method provides a new perspective for deepening the understanding of delay propagation mechanisms in airport networks.

CGKPN: Cross-Graph Knowledge Propagation Network with Adaptive Connection for Reasoning-Based Machine Reading Comprehension

The task of machine reading comprehension (MRC) is to enable machine to read and understand a piece of text and then answer the corresponding question correctly. This task requires machine to not only be able to perform semantic understanding but also possess logical reasoning capabilities. Just like human reading, it involves thinking about the text from two interacting perspectives of semantics and logic. However, previous methods based on reading comprehension either consider only the logical structure of the text or only the semantic structure of the text and cannot simultaneously balance semantic understanding and logical reasoning. This single form of reasoning cannot make the machine fully understand the meaning of the text. Additionally, the issue of sparsity in composition presents a significant challenge for models that rely on graph-based reasoning. To this end, a cross-graph knowledge propagation network (CGKPN) with adaptive connection is presented to address the above issues. The model first performs self-view node embedding on the constructed logical graph and semantic graph to update the representations of the graphs. Specifically, a relevance matrix between nodes is introduced to adaptively adjust node connections in response to the challenge posed by sparse graph. Subsequently, CGKPN conducts cross-graph knowledge propagation on nodes that are identical in both graphs, effectively resolving conflicts arising from identical nodes in different views, and enabling the model to better integrate the logical and semantic relationships of the text through efficient interaction. Experiments on the two MRC datasets ReClor and LogiQA indicate the superior performance of our proposed model CGKPN compared to other existing baselines.

Multilayer networks for measuring interconnectedness among global stock markets through the lens of trading volume-price relationship

Using daily data spanning from 5 January 2004 to 22 November 2022, we quantify the spillover effects between 42 global stock markets. Specifically, combining causal structure learning and Elastic-Net-VAR methods, we innovatively construct multilayer causal networks based on volume-price relationship. Then, we analyze the network characteristics of multilayer spillover networks from system and market levels. Our findings indicate that there is heterogeneity in risk spillovers of price and volume networks, highlighting how trading volume spillover network play an important role in the risk contagion. Furthermore, multilayer interconnected networks confirm the risk spillovers between volume and price, and it exhibits significant differences compared to single-layer network. In addition, at system-level, each network layer shows unique network structures and dynamic evolution characteristics. At market-level, global stock markets play different roles in emitting or receiving shocks through various transmission channels. Our study emphasizes the importance of intra- and inter-layer risk propagation in multilayer networks based on volume and price, and has significant implications for developing investment strategies and global portfolio risk management.