Network Links Research Articles

Nonlinear acoustic echo cancellation using low-complexity low-rank recursive least-squares algorithms

Adaptive exponential functional link networks (AEFLN) are a type of linear-in-the-parameter nonlinear filters, which have shown enhanced modeling capability for nonlinear systems. However, the convergence speed of traditional AEFLN is generally slow for long impulse responses, hence, AEFLN trained using recursive least-squares becomes an attractive choice to achieve faster convergence. Moreover, huge computational burden of RLS makes it unsuitable in practical applications like echo cancellation. While low complexity versions of RLS are widely available in literature, they still suffer from low convergence speed. To address this issue, we propose a nonlinear acoustic echo cancellation (NAEC) system using AEFLN-RLS, based on the nearest Kronecker product (NKP) decomposition and low-rank approximation technique, which not only reduces computational complexity but also achieves improved convergence speed (especially tracking). To further improve the echo cancellation performance in non-stationary conditions, a variable regularization approach based NKP-AEFLN-RLS system is also proposed. To also reduce the computational complexity further, dichotomous coordinate descent (DCD) updates are incorporated into the proposed NKP-AEFLN-RLS NAEC system and its variable regularization version. Experimental results show the effectiveness of the proposed algorithms, with the variable regularized version of the algorithm using DCD iterations showing the best compromise between convergence capability and lower computational complexity.

An efficient angle-based twin random vector functional link classifier

Random vector functional link (RVFL) has always proven to be an excellent classifier in various application areas of machine learning. In this work, inspired by RVFL and its twin variant, i.e., twin RVFL (TRVFL), we propose an efficient angle-based twin random vector functional link (ATRVFL) to address the binary classification problem. ATRVFL is a general classification model in which the first optimization problem can be found by solving a quadratic programming problem (QPP). The second problem can be described as an unconstrained minimization problem (UMP) which is solved by a system of linear equations. ATRVFL finds the solution by solving a QPP and a UMP instead of solving two QPPs. Therefore, its training time is significantly less than TRVFL. To maximize the angle between the normals of the two hyperplanes, the second hyperplane is chosen to be close to its class. The objective of introducing the idea of angle is to maximise the distance between two hyperplanes to improve the ability to discriminate them in geometric spaces. Experimental analyses are performed on a leaf dataset and 25 real-world benchmark datasets (RWDs) collected from the UCI repository of various fields like medical, biological, etc. The results are evaluated based on classification accuracy, the area under the curve, G-mean and F1-score. The results of ATRVFL are compared with Support Vector Machine (SVM), RVFL, Twin-RVFL (TRVFL), RVFL with ε-insensitive Huber loss (ε-HRVFL), ensemble deep learning-based RVFL network (edRVFL) and Intuitionistic fuzzy random vector functional link classifier (IFRVFLC). The classification accuracy of the proposed model on the leaf dataset is 91.093 % and the highest F1-score of 0.768. Overall, based on experimental analysis it is evident that the proposed ATRVFL shows comparative or better performance than SVM, RVFL, TRVFL, ε-HRVFL, IFRVFLC and edRVFL.

Investigating clinical links in edge-labeled citation networks of biomedical research: A translational science perspective

While clinical citations have been widely used as the preeminent measure of the clinical impact of biomedical paper, there has been a scarcity of in-depth studies exploring their temporal and structural characteristics, as well as its influence on the clinical translation. To fill this gap, we categorized biomedical papers and their citations into four groups from the translational science perspective: basic, clinical, mixed, and human-related. Subsequently, we constructed an edge-labeled citation network and four clinical translation networks. Our analysis encompassed 114,342 papers in the field of Alzheimer's Disease, accompanied by 5,161,626 citations, of which 2.77 % were clinical citations, 18.77 % basic citations, 41.85 % mixed citations, and 36.61 % human-related citations. First, utilizing time- and structure-randomized networks, we conducted a quantitative analysis of clinical citations' incidence patterns, impact assortativity, temporal occurrence patterns, and temporal co-location patterns throughout the lifecycles of biomedical research. Second, in comparison to control groups, we evaluated the short- and long-term impacts of different types of citations on the academic influence and clinical translation of biomedical research. Our findings reveal that clinical citations effectively bolster the academic influence of biomedical papers, and this positive effect appears to amplify over time. Conversely, while basic, mixed, and human-related citations may initially aid in the clinical translation of biomedical research, over 70 % of them exhibit an inhibitory effect on clinical translation in the long run. These findings afford us a deep and specific understanding of how clinical citations operate within the context of biomedical papers, thereby serving as a crucial guide for effectively promoting the clinical translation of biomedical research.

An Integrated Method for Cooperation Prediction in Complex Standard Networks

Standards play significant roles in the development of technology and economics, while the cooperation between drafters directly determines the quality of standard systems. The cooperation prediction is a significant while challenging problem for seeking new cooperation chances between drafting units due to their differences in experience and professional ability. In this study, an integrated artificial intelligence method is proposed for cooperation prediction using the link prediction method, text analysis, and network modeling. Specifically, we develop a multi-layer standard network formed by standard citation relationships and cooperation relationships between drafters. Then, a set of novel metrics is designed for predicting the cooperation between drafters considering the knowledge, experience, and professional capability. These metrics are further integrated into a neural network to improve the prediction accuracy. The priorities of our method in terms of prediction accuracy are verified with realistic data of Chinese environmental health standards. The prediction results provide strong support for the selection of drafters and further optimize the structure of standard systems.

Precise single step and multistep short-term photovoltaic parameters forecasting based on reduced deep convolutional stack autoencoder and minimum variance multikernel random vector functional network

Accurate prediction of photovoltaic (PV) power can significantly alleviate energy crises. However, the inherent randomness and intermittency of PV power pose challenges to the stability and safety of PV-penetrated grid systems. To address this, we have developed a novel hybrid model: a reduced deep convolutional stack autoencoder with a minimum variance multikernel random vector functional link network (RDCSAE-MVMRVFLN). This model enhances grid efficacy and safety. We extract the most informative band-limited intrinsic mode functions (BLIMFs) of highly nonlinear and nonstationary solar energy parameters using an entropy, kurtosis, and correlation coefficient-based information-oriented variational mode decomposition (IOVMD). These efficient BLIMFs are concatenated and input into the RDCSAE for rich, abstract, and discriminative representation computation. A less computationally complex MVMKRVFLN regression method, incorporating these refined representations, is proposed for superior prediction accuracy and reduced computational complexity. Our method shows exceptional performance in predicting solar temperature, irradiation, and power for multi-horizon forecasts with minimal error metrics (correlation coefficients of 0.999±0.001, 0.992±0.001, 0.986±0.02 and 0.978±0.02, and RMSE of 0.016±0.001, 0.024±0.001, 0.034±0.001 and 0.045±0.001 for the interval of 10 minutes, 30 minutes, 1 hour and 3 hours respectively) in both single-step and multi-step forecasting compared to conventional methods. The RDCSAE-MVMKRVFLN model is implemented on a high-speed Xilinx Virtex-5 FPGA embedded processor to validate its simplicity, robustness, and practicability. Additionally, we examine the prediction performance using real-time data from a 1 MW solar farm in Odisha, India, demonstrating the model’s effectiveness and superiority.

Parallel ensemble of a randomization-based online sequential neural network for classification problems using a frequency criterion

Randomization-based neural networks have gained wide acceptance in the scientific community owing to the simplicity of their algorithm and generalization capabilities. Random vector functional link (RVFL) networks and their variants are a class of randomization-based neural networks. RVFL networks have shown promising results in classification, regression, and clustering problems. For real-world applications, learning algorithms that can train with new samples over previous results are necessary because of to the constant generation of problems related to large-scale datasets. Various online sequential algorithms, commonly involving an initial learning phase followed by a sequential learning phase, have been proposed to address this issue. This paper presents a training algorithm based on multiple online sequential random vector functional link (OS-RVFL) networks for large-scale databases using a shared memory architecture. The training dataset is distributed among p OS-RVFL networks, which are trained in parallel using p threads. Subsequently, the test dataset samples are classified using each trained OS-RVFL network. Finally, a frequency criterion is applied to the results obtained from each OS-RVFL network to determine the final classification. Additionally, an equation was derived to reasonably predict the total training time of the proposed algorithm based on the learning time in the initial phase and the time scaling factor compared to the sequential learning phase. The results demonstrate a drastic reduction in training time because of data distribution and an improvement in accuracy because of the adoption of the frequency criterion.

Multitone Upconverted Phase-Modulated Signal RoF Link for 6G & Next-Generation Networks

The Next-Generation Networks offer a green environment with demanded data rates with ultra-low latency. Future efforts will be oriented in more degrees of freedom and system complexity with incorporating global optimization in terms of non-linearities, cost-effective, and reduced environmental impacts. The multitone upconverted radio over fibre (RoF) link with an optical modulator as phase modulator as desired alternatives with respect to optimization features and quantum photonic integration. In this paper, a framework has been developed to characterise the upconverted phase modulated signal model based on next-generation networks, in which an analytical model is investigated for single-tone upconverted, two-tone upconverted, and three-tone upconverted. Numerical simulations have been carried out to verify the novel outcomes using OptSIM software. The developed analytical models have a major contribution to the implementation of next-generation networks such as 5G, 6G, WiMAX, Quantum Communication, etc. The fundamentals of harmonics and the intermodulation product terms of third and fifth order via optical link are also evaluated for multitone Phase modulated RoF link with upconversion method. As per the finding outcomes, the proposed link is adaptable with frequency upconversion and long fibre distances.

A methodology for ranking of critical links in transportation networks based on criticality score distributions

Criticality of transportation networks links is calculated through utilizing criticality metric(s) for various disruption scenarios that generally include one link removal/degradation at a time or simultaneous removal/degradation of multiple links. As also shown in this paper, single link removals do not reveal the actual criticality of link(s) due to network dependencies. Multiple simultaneous link removal scenarios can capture the network interactions; however, the results relate to the criticality ranking of scenarios than the links, e.g., the links in scenario-1 are more critical than the links in scenario-2. This paper addressed the need for a methodology that provides individual link criticality rankings based on failure scenarios. The proposed approach utilizes the distribution of the criticality scores for each link based on running traffic assignment for each failure scenario, e.g., the distribution for link #X is composed of the criticality scores of all link failure combinations that include link #X. The criticality ranking for each link is identified based on its criticality distribution's mean, coefficient of variation and skewness. The use of the proposed approach is illustrated on test networks to show that it provides robust link criticality rankings that accounts for network interactions across varying scales of link failure scenarios.

Learning to route and schedule links in reconfigurable networks

This paper considers networks with a reconfigurable topology with so called 60 GHz dynamic links that can be activated or disabled over time. A fundamental problem is to jointly determine which 60 GHz dynamic links are active and the route chosen by source nodes over time. To this end, this paper outlines a hierarchical deep reinforcement learning solution that can be used to compute the optimal policy that determines for each time slot (i) active dynamic links, and (ii) the route used by each source–destination pair. The results show that the proposed approach results in a maximum average queue length that is 80% shorter than non-learning methods.

Combating temporal composition inference by high-order camouflaged network topology obfuscation

Topology inference driven by non-collaborative or incomplete prior knowledge is widely used in pivotal target network sieving and completion. However, perceivable topology also allows attackers to identify the fragile bottlenecks and perform efficacious attacks that are difficult to defend against by injecting indistinguishable low-volume attacks. Most existing countermeasures are proposed to obfuscate network data or set up honeypots with adversarial examples. However, there are two challenges when adding perturbations to live network links or nodes. Firstly, the perturbations imposed on the network cannot be conveniently projected to the original network with poor scalability. Secondly, applying significant changes to network information is laborious and impractical. In short, making a good trade-off between concealment and complexity is challenging. To address the above issues, we propose a fraudulent proactive defending tactic, namely HBB-TSP, to protect live network privacy by combating attacks of temporal network inference. Specifically, to penetrate the critical network structures, HBB-TSP first brings in the Statistical Validation of Hypergraph (SVH) method to identify the pivotal connection information of the network and extract the deep backbone structure. Then, the Temporal Simple Decomposition Weighting (TSDW) strategy is introduced, which can predict the backbone network with evolution rules and add highly obfuscated features at a minimized overhead. Finally, a discriminator with multiple centrality models is used to evaluate the deceptiveness and, in turn, affect the TSDW prediction. The entire process ensures the consistency and robustness of network changes while ensuring effective adversarial resistance. Experimental results on two scale real-world datasets demonstrate the effectiveness and generalization of adversarial perturbations. In particular, it is encouraging that our proposed defending scheme outperforms the advanced countermeasures. It ensures the realization of a deceptive obfuscated network at minimum overhead and is suitable for widespread deployment in scenarios of different scales.

Research on reliability mapping of 5G low orbit constellation network slice based on deep reinforcement learning

Reliability mapping of 5G low orbit constellation network slice is an important means to ensure link network communication. The problem of state space explosion is a typical problem. The deep reinforcement learning method is introduced. Under the 5G low orbit constellation integrated network architecture based on software definition network (SDN) and network function virtualization (NFV), the resource requirements and resource constraints of the virtual network function (VNF) are comprehensively considered to build the 5G low orbit constellation network slice reliability mapping model, and the reliability mapping model parameters are trained and learned by using deep reinforcement learning, solve the problem of state space explosion in the reliability mapping process of 5G low orbit constellation network slices. In addition, node backup and link backup strategies based on importance are adopted to solve the problem that VNF/link reliability is difficult to meet in the reliability mapping process of 5G low orbit constellation network slice. The experimental results show that this method improves the network throughput, packet loss rate and intra slice traffic of 5G low orbit constellation, and can completely repair network faults within 0.3 s; For different number of 5G low orbit constellation network slicing requests, the reliability of this method remains above 98%; For SFC with different lengths, the average network delay of this method is less than 0.15 s.

Graph Embedded Ensemble Deep Randomized Network for Diagnosis of Alzheimer's Disease.

Randomized shallow/deep neural networks with closed form solution avoid the shortcomings that exist in the back propagation (BP) based trained neural networks. Ensemble deep random vector functional link (edRVFL) network utilize the strength of two growing fields, i.e., deep learning and ensemble learning. However, edRVFL model doesn't consider the geometrical relationship of the data while calculating the final output parameters corresponding to each layer considered as base model. In the literature, graph embedded frameworks have been successfully used to describe the geometrical relationship within data. In this paper, we propose an extended graph embedded RVFL (EGERVFL) model that, unlike standard RVFL, employs both intrinsic and penalty subspace learning (SL) criteria under the graph embedded framework in its optimization process to calculate the model's output parameters. The proposed shallow EGERVFL model has only single hidden layer and hence, has less representation learning. Therefore, we further develop an ensemble deep EGERVFL (edEGERVFL) model that can be considered a variant of edRVFL model. Unlike edRVFL, the proposed edEGERVFL model solves graph embedded based optimization problem in each layer and hence, has better generalization performance than edRVFL model. We evaluated the proposed approaches for the diagnosis of Alzheimer's disease and furthermore on UCI datasets. The experimental results demonstrate that the proposed models perform better than baseline models. The source code of the proposed models is available at https://github.com/mtanveer1/.

Bridging aquatic invasive species threats across multiple sectors through One Biosecurity.

Understanding the magnitude of biosecurity risks in aquatic environments is increasingly complex and urgent because increasing volumes of international shipping, rising demand for aquaculture products, and growth in the global aquarium trade, are accelerating invasive alien species spread worldwide. These threats are especially pressing amid climate and biodiversity crises. However, global and national biosecurity systems are poorly prepared to respond because of fragmented research and policy environments, that often fail to account for risks across sectors or across stakeholder needs and fail to recognize similarities in the processes underpinning biological invasions. In the present article, we illustrate the complex network of links between biosecurity threats across human, animal, plant, and environment sectors and propose a universal approach to risk assessment. One Biosecurity is a holistic, interdisciplinary approach that minimizes biosecurity risks across human, animal, plant, algal, and ecosystem health and is critical to reduce redundancy and increase cross-sectoral cohesion to improve policy, management, and research in aquatic biosecurity.

Lesions Causing Alice in Wonderland Syndrome Map to a Common Brain Network Linking Body and Size Perception

ObjectiveAlice in Wonderland syndrome (AIWS) profoundly affects human perception of size and scale, particularly regarding one's own body and the environment. Its neuroanatomical basis has remained elusive, partly because brain lesions causing AIWS can occur in different brain regions. Here, we aimed to determine if brain lesions causing AIWS map to a distributed brain network.MethodsA retrospective case–control study analyzing 37 cases of lesion‐induced AIWS identified through systematic literature review was conducted. Using resting‐state functional connectome data from 1,000 healthy individuals, the whole‐brain connections of each lesion were estimated and contrasted with those from a control dataset comprising 1,073 lesions associated with 25 other neuropsychiatric syndromes. Additionally, connectivity findings from lesion‐induced AIWS cases were compared with functional neuroimaging results from 5 non‐lesional AIWS cases.ResultsAIWS‐associated lesions were located in various brain regions with minimal overlap (≤33%). However, the majority of lesions (≥85%) demonstrated shared connectivity to the right extrastriate body area, known to be selectively activated by viewing body part images, and the inferior parietal cortex, involved in size and scale judgements. This pattern was uniquely characteristic of AIWS when compared with other neuropsychiatric disorders (family‐wise error‐corrected p < 0.05) and consistent with functional neuroimaging observations in AIWS due to nonlesional causes (median correlation r = 0.56, interquartile range 0.24).InterpretationAIWS‐related perceptual distortions map to one common brain network, encompassing regions critical for body representation and size‐scale processing. These findings lend insight into the neuroanatomical localization of higher‐order perceptual functions, and may inform future therapeutic strategies for perceptual disorders. ANN NEUROL 2024;96:662–674

Ensemble Deep Random Vector Functional Link Network Using Privileged Information for Alzheimer's Disease Diagnosis.

Alzheimer's disease (AD) is a progressive brain disorder. Machine learning models have been proposed for the diagnosis of AD at early stage. Recently, deep learning architectures have received quite a lot attention. Most of the deep learning architectures suffer from the issues of local minima, slow convergence and sensitivity to learning rate. To overcome these issues, non-iterative learning based deep randomized models especially random vector functional link network (RVFL) with direct links have proven to be successful. However, deep RVFL and its ensemble models are trained only on normal samples. In this paper, deep RVFL and its ensembles are enabled to incorporate privileged information, as the standard RVFL model and its deep models are unable to use privileged information. To fill this gap, we have incorporated learning using privileged information (LUPI) in deep RVFL model, and propose deep RVFL with LUPI framework (dRVFL+). Privileged information is available while training the models. As RVFL is an unstable classifier, we propose ensemble deep RVFL+ with LUPI framework (edRVFL+) which exploits the LUPI as well as the diversity among the base leaners for better classification. Unlike traditional ensemble approach wherein multiple base learners are trained, the proposed edRVFL+ model optimises a single network and generates an ensemble via optimization at different levels of random projections of the data. Both dRVFL+ and edRVFL+ efficiently utilise the privileged information which results in better generalization performance. In LUPI framework, half of the available features are used as normal features and rest as the privileged features. However, we propose a novel approach for generating the privileged information. We utilise different activation functions while processing the normal and privileged information in the proposed deep architectures. To the best of our knowledge, this is first time that a separate privileged information is generated. The proposed dRVFL+ and edRVFL+ models are employed for the diagnosis of Alzheimer's disease. Experimental results demonstrate the superiority of the proposed dRVFL+ and edRVFL+ models over baseline models. Thus, the proposed edRVFL+ model can be utilised in clinical setting for the diagnosis of AD.

Regional traffic congestion coordination control based on critical links

Critical links have the characteristics of carrying large traffic flow and quickly affecting the operational performance of the road network. Existing regional traffic congestion coordination control methods do not consider the impact of critical links. To solve this issue, this paper proposes a regional traffic congestion coordination control method based on critical links. First, the median, gravitational index, and traffic sharing ratio are selected as the identification indexes of critical links in urban networks. Second, the road network vulnerability is assessed, and road link criticality is calculated based on joint entropy theory and multi-scale factor synthesis of multi-metric identification results. Then, the regional signal control range is determined, and a multi-stage signal control optimization model is developed. Finally, the optimal traffic signals for intersections are obtained by minimizing the pressure at the maximum phase of the road link and maximizing the capacity of critical links. To demonstrate the effectiveness of the proposed methods, two benchmarks (i.e., the MAXBAND method and existing traffic signals) are compared. The results show that (1) in terms of boundary control, the average vehicle delay at the boundary intersection of the road network is 34.62 s, an increase of 7.49 s and 5.03 s compared to the benchmark methods, respectively. (2) In terms of internal control, the average delay of the road network based on the proposed method is 87.54 s, which is reduced by 16.38 % and 28.73 % compared with the benchmark methods, and the queue lengths of the critical links are reduced by 21.47 % and 34.15 %, respectively. (3) The control process of the proposed method enables the number of vehicles within the road network to be maintained near the optimal carrying capacity, which gives the road network an efficient operating efficiency.

Revisiting the evolution of bow-tie architecture in signaling networks

Bow-tie architecture is a layered network structure that has a narrow middle layer with multiple inputs and outputs. Such structures are widely seen in the molecular networks in cells, suggesting that a universal evolutionary mechanism underlies the emergence of bow-tie architecture. The previous theoretical studies have implemented evolutionary simulations of the feedforward network to satisfy a given input-output goal and proposed that the bow-tie architecture emerges when the ideal input-output relation is given as a rank-deficient matrix with mutations in network link intensities in a multiplicative manner. Here, we report that the bow-tie network inevitably appears when the link intensities representing molecular interactions are small at the initial condition of the evolutionary simulation, regardless of the rank of the goal matrix. Our dynamical system analysis clarifies the mechanisms underlying the emergence of the bow-tie structure. Further, we demonstrate that the increase in the input-output matrix reduces the width of the middle layer, resulting in the emergence of bow-tie architecture, even when evolution starts from large link intensities. Our data suggest that bow-tie architecture emerges as a side effect of evolution rather than as a result of evolutionary adaptation.

A new relaying approach for protecting TCSC compensated transmission line connected to DFIG based wind farm

In modern power systems, Thyristor Controlled Series Capacitor (TCSC)-compensated transmission lines are crucial in transporting bulk power from large wind farms. However, the performance of conventional distance relays is adversely impacted by the typical fault characteristics of the Doubly Fed Induction Generator (DFIG) and TCSC. To address this challenge, this paper presents a new relaying algorithm such as Complete Ensemble Empirical Mode Decomposition with Adaptive Noise-based Dominant Mode Algorithm-Hilbert Transform (CEEMDAN-DMA-HT) for fault detection and CEEMDAN assisted Enhanced Jaya Optimization-based Random Vector Functional Link Network (EJAYA-RVFL) for fault classification. In proposed fault detection algorithm, the differential current from both ends of the line is subjected to Complete Ensemble Empirical Mode Decomposition with Adaptive Noise, leading to the extraction of distinct intrinsic mode functions (IMFs). Employing the Dominant Mode Algorithm, the IMF with the highest Pearson correlation coefficient, referred to as the dominant IMF, is identified. Subsequently, a comparison between this dominant mode IMF and the original input signal is conducted after passing the obtained signal through the Hilbert Transform (HT), enabling effective fault detection. The proposed classifier leverages a straightforward feature set that quantifies the energy of each phase and the ground. These features serve as inputs for the proposed classifier, facilitating the categorization of different fault types. To evaluate the effectiveness of the proposed relaying algorithm, several fault and non-fault scenarios are simulated on a test power system using MATLAB/Simulink. The results demonstrate that; the proposed fault detection and classification algorithm are able to detect and classify faults in less than a half-cycle. The proposed method gives 100% fault accuracy for fault detection and 99.97 % accuracy for fault classification. Furthermore, the proposed classifier achieves the performance metrics like Precision (0.998), Recall (0.997) and F1-Score (0.996), providing quantitative insights into its accuracy and dependability. Finally, a comparative analysis is carried out with existing approaches.

A model-less approach for the optimal coordination of renewable energy sources and DC links in low-voltage distribution networks

This paper presents a model-less centralized control approach for coordinating Renewable Energy Sources and multiterminal DC links in low-voltage distribution networks. The controller is based on the Online Feedback Optimization technique which is adequate for this part of the power system, characterized by a lack of precise model and real-time information. The paper includes the complete mathematical formulation of the optimization problem and its solution applying the presented strategy. The performance of the proposed centralized controller is evaluated through simulations in the European LV distribution network proposed by the CIGRE Task Force C06.04.02. The results show the benefits that DC links may bring to LV networks due to the release of their radial operation. In particular, the proposed strategy achieves a 7.4 % daily energy loss reduction with respect to the base case. The centralized controller, which operates without a detailed network model and reduced real-time information, enables the use of this technology in LV networks which maximizes the penetration of renewable energy sources.

Supervised learning to covering cost risk through post-construction evaluation of transportation projects by project delivery methods

PurposeThe main objectives of this study are to (1) develop and test a cost contingency learning model that can generalize initially estimated contingency amounts by analyzing back the multiple project changes experienced and (2) uncover the hidden link of the learning networks using a curve-fitting technique for the post-construction evaluation of cost contingency amounts to cover cost risk for future projects.Design/methodology/approachBased on a total of 1,434 datapoints collected from DBB and DB transportation projects, a post-construction cost contingency learning model was developed using feedforward neural networks (FNNs). The developed model generalizes cost contingencies under two different project delivery methods (i.e. DBB and DB). The learning outputs of generalized contingency amounts were curve-fitted with the post-construction schedule and cost information, specifically aiming at uncovering the hidden link of the FNNs. Two different bridge projects completed under DBB and DB were employed as illustrative examples to demonstrate how the proposed modeling framework could be implemented.FindingsWith zero or negative values of change growth experienced, it was concluded that cost contingencies were overallocated at the contract stage. On the other hand, with positive values of change growth experienced, it was evaluated that set cost contingencies were insufficient from the post-construction standpoint. Taken together, this study proposed a tangible post-construction evaluation technique that can produce not only the plausible ranges of cost contingencies but also the exact amounts of contingency under DBB and DB contracts.Originality/valueAs the first of its kind, the proposed modeling framework provides agency engineers and decision-makers with tangible assessments of cost contingency coupled with experienced risks at the post-construction stage. Use of the proposed model will help them evaluate the allocation of appropriate contingency amounts. If an agency allocates a cost contingency benchmarked from similar projects on aspects of the base estimate and experienced risks, a set contingency can be defended more reliably. The main findings of this study contribute to post-construction cost contingency verification, enabling agency engineers and decision-makers to systematically evaluate set cost contingencies during the post-construction assessment stage and achieving further any enhanced level of confidence for future cost contingency plans.