Low Discrepancy Research Articles

ПІДВИЩЕННЯ ГАЛЬМІВНОЇ ЗДАТНОСТІ ЛОКОМОТИВА ПРОМИСЛО-ВОГО ТРАНСПОРТУ З УРАХУВАННЯМ ЗУСИЛЬ У ЙОГО ЗЧІПНИХ ПРИСТРОЯХ

Objective: substantiation the rational concept of increasing the braking capacity of the industrial locomotive in transient modes of train movement, taking into account the force factor in its coupling device. Methodology of the work: theoretical studies of the energy from the deformation of the coupling devices of the industrial locomotive in the transient modes of train movement; theoretical studies of excessive coupling of locomotive wheels with rails in the presence of force on the locomotive's coupling device; determination of kinematic and dynamic parameters of the transient mode of train movement. A mathematical model of the train motion transitory process has been developed, it using the example of locomotive braking process, which allows determining the current values of train element velocities, as well as coupling forces, and has a low average discrepancy with the data of the technological calculation of industrial locomotive. Fixed: the force on the coupling device of the locomotive changes with a high frequency. The braking process is conventionally divided into 3 stages, which are characterized by oscillations, stabilization and a drop in clutch force; the energy corresponding to the change in force in the clutch during braking falls on 90.5% of the first and second stages. Taking into account the obtained results, the conceptual adjustment of the force of the coupling of the wheel pairs of the locomotive with the rails at the level of the maximum forces in the coupling device of the locomotive, determined by the simulation results, is adopted. Scientific novelty: improvement of the method of determining the required force of the traction the wheels of locomotive with the rails by taking into account the energy corresponding to the change in the effort in the coupling device of the locomotive during the braking time of the train. Practical value: the development of an improved concept by regulating the force the traction of locomotive wheel pairs with rails in transient modes of train movement, taking into account the energy assessment of the effort in the locomotive's coupling device.

Message-Passing Monte Carlo: Generating low-discrepancy point sets via graph neural networks

Discrepancy is a well-known measure for the irregularity of the distribution of a point set. Point sets with small discrepancy are called low discrepancy and are known to efficiently fill the space in a uniform manner. Low-discrepancy points play a central role in many problems in science and engineering, including numerical integration, computer vision, machine perception, computer graphics, machine learning, and simulation. In this work, we present a machine learning approach to generate a new class of low-discrepancy point sets named Message-Passing Monte Carlo (MPMC) points. Motivated by the geometric nature of generating low-discrepancy point sets, we leverage tools from Geometric Deep Learning and base our model on graph neural networks. We further provide an extension of our framework to higher dimensions, which flexibly allows the generation of custom-made points that emphasize the uniformity in specific dimensions that are primarily important for the particular problem at hand. Finally, we demonstrate that our proposed model achieves state-of-the-art performance superior to previous methods by a significant margin. In fact, MPMC points are empirically shown to be either optimal or near-optimal with respect to the discrepancy for low dimension and small number of points, i.e., for which the optimal discrepancy can be determined.

Evaluation of industrial and consumer 3-D resin printer fabrication of microdevices for quality management of genetic resources in aquatic species

Aquatic germplasm repositories can play a pivotal role in securing the genetic diversity of natural populations and agriculturally important aquatic species. However, existing technologies for repository development and operation face challenges in terms of accuracy, precision, efficiency, and cost-effectiveness, especially for microdevices used in gamete quality evaluation. Quality management is critical throughout genetic resource protection processes from sample collection to final usage. In this study, we examined the potential of using three-dimensional (3-D) stereolithography resin printing to address these challenges and evaluated the overall capabilities and limitations of a representative industrial 3-D resin printer with a price of US$18,000, a consumer-level printer with a price <US$700, and soft lithography, a conventional microfabrication method. A standardized test object, the Integrated Geometry Sampler (IGS), and a device with application in repository quality management, the Single-piece Sperm Counting Chamber (SSCC), were printed to determine capabilities and evaluate differences in targeted versus printed depths and heights. The IGS design had an array of negative and positive features with dimensions ranging from 1 mm to 0.02 mm in width and depth. The SSCC consisted of grid and wall features to facilitate cell counting. The SSCC was evaluated with polydimethylsiloxane (PDMS) devices cast from a typical photoresist and silicon mold. Fabrication quality was evaluated by optical profilometry for parameters such as dimensional accuracy, precision, and visual morphology. Fabrication time and cost were also evaluated. The precision, reliability, and surface quality of industrial-grade 3-D resin printing were satisfactory for operations requiring depths or heights larger than 0.1 mm due to a low discrepancy between targeted and measured dimensions across a range of 1 mm to 0.1 mm. Meanwhile, consumer-grade printers were suitable for microdevices with depths or heights larger than 0.2 mm. While the performance of either of these printers could be further optimized, their current capabilities, broad availability, low cost of operation, high throughput, and simplicity offer great promise for rapid development and widespread use of standardized microdevices for numerous applications, including gamete quality evaluation and “laboratory-on-a-chip” applications in support of aquatic germplasm repositories.

Solar receiver endurance assessment under non-conventional operation modes

Central receiver endurance in solar power tower plants is a critical aspect to ensure their viability. In this work, two receiver configurations, with and without flow path crossover, are compared in thermal and mechanical performance. Considering their thermomechanical limits during operation, results show that panels in the crossover configuration would endure from 1.22 to 2 times that in the no-crossover, just marginally penalising its thermal efficiency. Additionally, two plant operation strategies are compared: continuous operation and nighttime strategy. The former results in 1.89-times more damage but roughly a 1.9-times increase in energy output, obtaining a practically constant trade-off between damage and energy generation for these two strategies. Lastly, the effect of considering either a linear or bilinear (interaction point set at 0.05,0.05) creep/fatigue damage interaction is explored. While producing more conservative results, linear interaction error is below 3 % in the most critical receiver regions. Such a low discrepancy is caused by the creep-damage dominance over fatigue in the most critical regions. Other areas show heavier creep-fatigue interaction, resulting in larger error, but these areas obtain times to failure much longer than expected receiver lifespan.

A deep neural network reduced order model for unsteady aerodynamics of pitching airfoils

A machine learning framework is developed to compute the aerodynamic forces and moment coefficients for a pitching NACA0012 airfoil incurring in light and deep dynamic stall. Four deep neural network frameworks of increasing complexity are investigated: two multilayer perceptrons and two convolutional neural networks. The convolutional framework, in addition to the standard mean squared error loss, features an improved loss function to compute the airfoil loads. In total, five models are investigated of increasingly complexity. The convolutional model, coupled with the loss function based on force and moment coefficients and embedding the attention mechanism, is found to robustly and efficiently predict pressure and skin friction distributions over the airfoil over the entire pitching cycle. Periodic conditions are implemented to grant the physical smoothness of the model output both in space and time. An analysis of the training dataset point distributions is performed to point out the effects of adopting low discrepancy sequences, such as Latin hypercube, Sobol', and Halton, compared to random and uniform sequences. The current model shows improved performances in predicting forces and pitching moment in a broad range of operating conditions.

Quasi-random Fractal Search (QRFS): A dynamic metaheuristic with sigmoid population decrement for global optimization

Global optimization of complex and high-dimensional functions remains a central challenge with broad applications in science and engineering. This study introduces a new optimization approach called quasi-random metaheuristic based on fractal search (QRFS), which harnesses the power of fractal geometry, low discrepancy sequences, and intelligent search space partitioning techniques. The QRFS uses fractals’ inherent self-similarity and intricate structure to guide the solution space exploration. For the proposal, a deterministic but quasi-random element is used in the search process using low discrepancy sequences, such as Sobol, Halton, Hammersley, and Latin Hypercube. This integration allows the algorithm to systematically cover the search space while maintaining the level of diversity necessary for efficient exploration. The QRFS employs a dynamic strategy of partitioning the search space and reducing the population of solutions to optimize the use of function accesses, which causes it to adapt well to the characteristics of the problem. The algorithm intelligently identifies and prioritizes promising regions within the fractal-based representation, allocating computational resources where they are most likely to yield optimal solutions. Experimental evaluations on several benchmark problems demonstrate that QRFS consistently outperforms modern, canonical metaheuristics and variants of algorithms such as differential evolution (DE), particle swarm optimization (PSO), covariance matrix adaptive evolution strategy (CMA-ES), regarding solution quality. Besides, the algorithm shows remarkable scalability, which makes it suitable for high-dimensional optimization tasks. Overall, QRFS offers a robust and efficient approach to solving complex optimization problems in various domains, paving the way for improved decision-making in real-world applications.

Investigating embedded data distribution strategy on reconstruction accuracy of flow field around the crosswind-affected train based on physics-informed neural networks

PurposePhysics-informed neural networks (PINNs) have become a new tendency in flow simulation, because of their self-advantage of integrating both physical and monitored information of fields in solving the Navier–Stokes equation and its variants. In view of the strengths of PINN, this study aims to investigate the impact of spatially embedded data distribution on the flow field results around the train in the crosswind environment reconstructed by PINN.Design/methodology/approachPINN can integrate data residuals with physical residuals into the loss function to train its parameters, allowing it to approximate the solution of the governing equations. In addition, with the aid of labelled training data, PINN can also incorporate the real site information of the flow field in model training. In light of this, the PINN model is adopted to reconstruct a two-dimensional time-averaged flow field around a train under crosswinds in the spatial domain with the aid of sparse flow field data, and the prediction results are compared with the reference results obtained from numerical modelling.FindingsThe prediction results from PINN results demonstrated a low discrepancy with those obtained from numerical simulations. The results of this study indicate that a threshold of the spatial embedded data density exists, in both the near wall and far wall areas on the train’s leeward side, as well as the near train surface area. In other words, a negative effect on the PINN reconstruction accuracy will emerge if the spatial embedded data density exceeds or slips below the threshold. Also, the optimum arrangement of the spatial embedded data in reconstructing the flow field of the train in crosswinds is obtained in this work.Originality/valueIn this work, a strategy of reconstructing the time-averaged flow field of the train under crosswind conditions is proposed based on the physics-informed data-driven method, which enhances the scope of neural network applications. In addition, for the flow field reconstruction, the effect of spatial embedded data arrangement in PINN is compared to improve its accuracy.

HB-RRT:A path planning algorithm for mobile robots using Halton sequence-based rapidly-exploring random tree

Path planning remains crucial for efficient robot operation. A Halton Biased Rapidly-exploring Random Tree (HB-RRT) path planning algorithm is introduced in this study. The Halton sequence, known for its uniform distribution and low discrepancy, is employed for sampling. Issues arising from the pseudo-random sequence in the standard RRT algorithm, leading to uneven distribution of sampling points, are addressed. A mouse-inspired goal-oriented strategy and a candidate sampling pool strategy are incorporated to enhance the sampling point quality, thereby addressing the challenge of insufficient memory during node expansion. Path optimization is further achieved through a multi-level planning approach, which aims to minimize redundancy. A subsequent smoothing of the path is conducted using a cubic B-spline method. Comparisons with the RRT, Bionic Target Bias-RRT, and Informed-RRT* algorithms, through both numerical simulations and real-world testing, confirm the superiority of the HB-RRT algorithm in terms of planning time, path length, and overall path quality.

A comparison of uterine corpus and endometrium volumes measured in 2D and 3D modes

Objective: to compare the uterine corpus and endometrium volumes measured in 2D and 3D modes.Material and methods. The observational retrospective cohort study included 154 women of reproductive age with no myometrial or endometrial pathology. Pelvic ultrasound was performed with the use of the Affiniti70 (Philips, Netherlands) with a multifrequency 3D intracavitary probe. The uterine corpus volume and endometrial volume were measured both in 2D and 3D modes, followed by a calculation of the percentage ratio of endometrial volume to uterine corpus volume (endometrial/uterine corporeal volume ratio (EV/UCV)).Results. The values of uterine corpus volume measured in 3D mode were higher than in 2D mode, with a relative measurement error of 7.2%. The strength of the correlation turned out to be very high (r = 0.91, p = 0.458). According to the Bland-Altman plot, almost all values of the volume difference in pairwise measurements fell within the interval ±1.96 SD 95%; a low average difference indicates a low systematic discrepancy in measurements, and the degree of value scatter is quite acceptable. The values of endometrium volume in 3D mode were lower than in 2D mode; the relative error in 2D mode, regardless of the cycle phase, was -35.3%. There was a strong correlation between the two measurement methods (r = 0.81), but the differences in allocations were significant (p&lt;0.05).Conclusion. It is permissible to use the values of the uterine corpus volume obtained in 2D mode as an analogue of 3D mode volume in routine practice, while it is not acceptable in the assessment of endometrium volume and EV/UCV.

Scenario analysis of local storylines to represent uncertainty in complex human-water systems

Storylines are important in evaluating the uncertainty inherent in complex human-water systems. The interrelated nature of qualitative and quantitative scenarios can enhance our ability to address the uncertainty of integrated modelling of complex systems. This study proposes a transdisciplinary approach that integrates social and environmental sciences to characterize and comprehend uncertainty in the dynamic interactions of key factors affecting a human-water system. We introduce a framework for representing uncertainty through linguistic and epistemic uncertainty quantification using storyline narratives in the context of a regional integrated dynamic model. A systematic exploration of uncertainty space is performed using storytelling, fuzzy sets, and low discrepancy sequences sampling methods. Scenario analysis is applied to the generated uncertain ensemble of projections to discover predominant storylines of interest. As a representative case of a human-water system operating in a developing country, we examine the uncertainty effects of a variety of drivers of climatic and socio-economic changes on key agriculture and water-related sectors in Pakistan’s Rechna Doab region. The findings revealed soil salinity and crop yield indices were the most uncertain and showed significant variance across all developed storylines. The 95th percentile for soil salinity in year 2100 was estimated to be nearly 60 % higher than the baseline level (year 2020). There was, however, considerable overlap in different socio-economic scenarios at the local scale, indicating that change in socio-economic conditions could not fully offset climate-related uncertainty. Our analysis provides better quantification and a deeper understanding of the uncertainty in integrated assessment modelling of coupled human-water systems and the complex relationships between inputs and outcomes.

Low discrepancy on non-linear sensor deployment in a time-critical linear IIoT network

Linear networks are characterized by their long and narrow topology, which transmits data in a linear manner. For time-critical linear IIoT networks, stable connectivity is paramount. A good deployment strategy ensures stable connectivity in the network and maximize the lifetime and coverage of the network. In order to achieve this goal, deployment strategies should address tunnelling effects in linear networks. The proposed method tackles three major issues: 1. minimizing tunnelling effects using non-linear sensor deployment method, 2. maximizing network coverage by introducing a modified Latin Hypercube Sequence (LHS), and 3. Minimizing data interference in the multichannel TSCH network using a neighbour count-based sensor node deployment technique. The simulation validates the efficacy of the suggested approach in prolonging the network’s lifespan by mitigating the tunnelling effect near the sink. Moreover, it optimally utilizes over 70% of the mean energy within the network system. This is a significant improvement over previous algorithms which often failed to even utilizing 40% of the energy, and many a times much lesser. In terms of network coverage, it surpasses other algorithms. Consequently, the proposed method delivers economic advantages by extending the network’s longevity, achieved through a higher number of nodes concentrated around the sink, and significantly reduces the tunnelling effect in the network.

Affective associations towards running: fuzzy patterns of implicit-explicit interaction in young female runners and non-runners.

Empirical evidence demonstrates that high concordance and low discrepancy of implicit and explicit affective processes facilitate consistent exercise behavior. Novice runners often have difficulties implementing their running behavior on a regular basis resulting in irregular running behavior. To investigate the potential value of affective associations 89 young female runners (regular and irregular) and non-runners were recruited. Affective associations towards running were measured through a Single-Target Implicit Association Test on the implicit level and by self-report on the explicit level. Implicit-explicit interaction (IEI) scores (i.e., implicit-explicit concordance and discrepancy) were derived from principal component analysis. Fuzzy k-means cluster analysis was used to identify patterns of interacting implicit-explicit affective associations. The resulting clusters were assessed for differences in previous running experience, current running behavior, motivational and intentional aspects. Four meaningful overlapping clusters were found and labeled according to their prevalent IEI patterns (i.e., "positive non-discrepant", "positive discrepant", "negative discrepant", "negative non-discrepant"). Significant differences between clusters were found for past running experience, current running behavior, motivational and intentional aspects. The results indicate that running behavior varies between and within patterns of affective associations. In line with previous findings, positive non-discrepant implicit and explicit affective associations are linked to more consistent running behavior, while negative non-discrepant affect is associated with non-runners. However, the occurrence of discrepant implicit-explicit affective associations in young women differing in running behavior, motivation, and intention broadens the view of the complex relationship between affective processes and exercise behavior. In conclusion, individualized interventions that take into account the implicit-explicit interaction of affective associations besides well-known cognitive self-regulatory resources may prove more effective for individuals who struggle to run regularly.

High-throughput determination of grain size distributions by EBSD with low-discrepancy sampling.

Because microstructure plays an important role in the mechanical properties of structural materials, developing the capability to quantify microstructures rapidly is important to enabling high-throughput screening of structural materials. Electron backscatter diffraction (EBSD) is a common method for studying microstructures and extracting information such as grain size distributions (GSDs), but is not particularly fast and thus could be a bottleneck in high-throughput systems. One approach to accelerating EBSD is to reduce the number of points that must be scanned. In this work, we describe an iterative method for reducing the number of scan points needed to measure GSDs using incremental low-discrepancy sampling, including on-the-fly grain size calculations and a convergence test for the resulting GSD based on the Kolmogorov-Smirnov test. We demonstrate this method on five real EBSD maps collected from magnesium AZ31B specimens and compare the effectiveness of sampling according to two different low discrepancy sequences, the Sobol and R2 sequences, and random sampling. We find that R2 sampling is able to produce GSDs that are statistically very similar to the GSDs of the full density grids using, on average, only 52% of the total scan points. For EBSD maps that contained monodisperse GSDs and over 1000 grains, R2 sampling only required an average of 39% of the total EBSDpoints.

An efficient quasi-Monte Carlo method with forced fixed detection for photon scatter simulation in CT.

Detected scattered photons can cause cupping and streak artifacts, significantly degrading the quality of CT images. For fast and accurate estimation of scatter intensities resulting from photon interactions with a phantom, we first transform the path probability of photons interacting with the phantom into a high-dimensional integral. Secondly, we develope a new efficient algorithm called gQMCFFD, which combines graphics processing unit(GPU)-based quasi-Monte Carlo (QMC) with forced fixed detection to approximate this integral. QMC uses low discrepancy sequences for simulation and is deterministic versions of Monte Carlo. Numerical experiments show that the results are in excellent agreement and the efficiency improvement factors are 4 ∼ 46 times in all simulations by gQMCFFD with comparison to GPU-based Monte Carlo methods. And by combining gQMCFFD with sparse matrix method, the simulation time is reduced to 2 seconds in a single projection angle and the relative difference is 3.53%.

Predicting Building Energy Demand and Retrofit Potentials Using New Climatic Stress Indices and Curves

Building energy requalification in Italy and Europe has been much discussed in recent years due to the high percentage of existing buildings with poor energy performance. In this context, it is useful to obtain a user-friendly and fast tool to predict the thermal energy demand (TED) for space conditioning and the related primary energy consumption (PEC) as a function of climatic stress. In this study, the SLABE methodology (simulation-based large-scale uncertainty/sensitivity analysis of building energy performance) is used to simulate representative Italian buildings, varying parameters such as geometry, envelope and HVAC (heating, ventilating and space conditioning) systems. MATLAB® in combination with EnergyPlus is used to analyze 200 buildings belonging to two structural types (multi-family buildings and apartment blocks) built in 1961–1975. Nine scenarios (as-built scenarios and eight retrofit ones) are investigated in 63 climatic locations. A regression analysis shows that the classical HDDs (heating degree days) approach cannot give an accurate prediction of TED because solar radiation is not accounted for. Thus, new climatic indices are developed alongside solar radiation, including the heating stress index (HSI), the cooling stress index (CSI) and the yearly climatic stress index (YCSI). The purpose of our work is to obtain climatic stress curves for the prediction of TED and PEC. Testing of this novel approach is performed by comparison with another building energy simulation tool, showing a low discrepancy, i.e., the coefficient of variation of the root mean square error is between 12% and 28%, which confirms certain reliability of the approach here proposed.

Investment certificates pricing using a Quasi-Monte Carlo framework: Case-studies based on the Italian market

The Monte Carlo method, thanks to its flexibility in designing even extremely complex payoffs, is assuming an increasingly important role in quantitative analysis. Its main limitation is the high computational cost linked to its modest speed of convergence to the fair value of the product. One of the best-known statistical techniques is to replace the random number generator with “low discrepancy” deterministic numerical sequences, producing a Quasi-Monte Carlo. Through its implementation for the analysis of three investment certificates featuring different characteristics and different stochastic processes used for the underlying simulation, the study demonstrates the possibility of achieving interesting results in terms of performance even for pricing these structured products ever more popular in the financial industry.

Time-efficient soft computational approach on vibration response of carbon-enriched drive shafts of super-fast race cars

Carbon-enriched nanocomposite drive shafts have received considerable attention in super-fast race cars because of their superb material and mechanical properties. thus, for the first time, vibration behavior of drive shafts enriched by three conventional distribution patterns of carbon nanotubes (CNT) in the polymer matrix are explored and reported. A new form in the framework of First-order Timoshenko beam theory is employed. The elasticity moduli of the CNT enriched nanocomposite material are approximated through a refined model of rule of mixtures. Hamilton’s variational principle serves as a way to determine the beam’s equations of motion. Differential quadrature approach (DQA) is employed to predict the system’s frequency for some conditions labeled as the training information. A new soft computation approach based on deep-learning methods employs the training information to adjust a predictor mechanism for obtaining the system’s vibration for any other conditions. This novel methodology provides this opportunity to considerably reduce the computational effort. Validity of the mentioned solution is verified through a comparative scrutinization by low discrepancy between its results and those of the published studies. Importance of these agility in computations would be discerned in cataloging process of the manufactured drive shafts in car industry.

Cross-cultural adaptation and validation of the Spanish version of the Oswestry disability index for Mexican population

Purpose This study aimed to adapt a Spanish translation of the Oswestry Disability Index (ODI) into a cross-cultural version for the Mexican population. The objectives were to verify the validity and reliability of the adapted ODI and to compare pain perception between patients with and without obesity. Material and methods We included 102 patients with low back pain from two neurosurgery departments in Mexico. The ODI questionnaire was translated and culturally adapted. Validity and construct were evaluated using exploratory factor analysis, and the external convergent validity was assessed by correlating ODI scores with pain intensity, age, and obesity. Test-retest reliability was calculated using the intraclass correlation coefficient, and confirmatory analysis was employed to validate the factorial structure. Results Patients with obesity were older and had higher pain scores than patients without obesity. The exploratory analysis of the ODI in Mexican Spanish showed good reliability (Cronbach’s alpha of 0.923) and validity (factorial loading range, 0.681 − 0.818). The confirmatory analysis showed almost null or very low discrepancy between the proposed model and the real data. Conclusions A Spanish translation of ODI was cross-culturally adapted for the Mexican population. The Mexican version of the ODI showed good reliability and validity in Mexican culture.

Biocomposites based on poly(lactic acid)/Aloe Vera/Priplast. Effects of UVA radiation on the crystallization

AbstractBiocomposites based on poly(lactic acid) (PLA), Aloe Vera (AV), and Priplast (Pr), were produced with AV and Pr levels of 5% and 10%, and submitted to UVA radiation for a time interval up to 28 days. Chemical changes were investigated using Fourier transform infrared spectroscopy (FTIR) and the thermal behavior was elucidated using differential scanning calorimetry (DSC); the non‐isothermal crystallization kinetics was modeled using the Pseudo‐Avrami model. FTIR spectra evidenced the formation of chromophore groups resulting from the photodegradation, which it was less intense in biocomposites with 10% Pr. AV addition promoted the cold crystallization, whereas the crystallization rate () displayed higher values for PLA/10%AV, with slight increases in the degree of crystallinity (Xc); while Pr addition decreased the Xc and of PLA, and protected it more intensely against the photodegradation effects. Pseudo‐Avrami model presented R2 ≥ 0.9909 and a low discrepancy between experimental and theoretical data, being adequate to describe the cold crystallization of PLA biocomposites. Summing up, AV and Pr addition protect PLA against the harmful effects of photodegradation, with AV subtly interfering in the crystalline parameters while Pr provides greater amorphicity to PLA biocomposites.

A complementary and contrastive network for stimulus segmentation and generalization

SummaryExisting convolutional neural networks (CNNs) have achieved remarkable performance in medical image segmentation tasks. However, they still fail to generalize well to unseen datasets due to the limited size and diversity of training data as well as distribution shifts. Meanwhile, CNN-based methods have inherent limitations in capturing global contexts and suffer semantic dilution issues in the decoder stage, which leads to suboptimal predictions especially under low inter-class discrepancy and complex backgrounds. In this paper, we propose a novel framework named CCNet that learns complementary and contrastive features for accurate segmentation. Firstly, a novel complementary feature extraction module is formulated to learn global–local features by coordinating Transformer and CNN-style parallel branches. Secondly, a global context refinement module is constructed to adaptively generate a set of layer-specific global maps, so as to remedy semantic dilution. Thirdly, a mutual attentive module is designed to alleviate background confusion, in which contrastive cues are mutually captured from the foreground and background view by cascaded dual attention blocks. Moreover, we implement synthetic data augmentation to deal with training data scarcity and distribution shifts, thereby improving the out-of-distribution generalization of our model. Extensive experiments demonstrate that our CCNet achieves outstanding performance in polyp, skin lesion, and nuclei segmentation tasks, outperforming the state-of-the-arts.