Nature Of Noise Research Articles

PatchNAS: Repairing DNNs in Deployment with Patched Network Architecture Search

Despite being widely deployed in safety-critical applications such as autonomous driving and health care, deep neural networks (DNNs) still suffer from non-negligible reliability issues. Numerous works had reported that DNNs were vulnerable to either natural environmental noises or man-made adversarial noises. How to repair DNNs in deployment with noisy samples is a crucial topic for the robustness of neural networks. While many network repairing methods based on data argumentation and weight adjustment have been proposed, they require retraining and redeploying the whole model, which causes high overhead and is infeasible for varying faulty cases on different deployment environments. In this paper, we propose a novel network repairing framework called PatchNAS from the architecture perspective, where we freeze the pretrained DNNs and introduce a small patch network to deal with failure samples at runtime. PatchNAS introduces a novel network instrumentation method to determine the faulty stage of the network structure given the collected failure samples. A small patch network structure is searched unsupervisedly using neural architecture search (NAS) technique with data samples from deployment environment. The patch network repairs the DNNs by correcting the output feature maps of the faulty stage, which helps to maintain network performance on normal samples and enhance robustness in noisy environments. Extensive experiments based on several DNNs across 15 types of natural noises show that the proposed PatchNAS outperforms the state-of-the-arts with significant performance improvement as well as much lower deployment overhead.

Anthropogenic noise decreases activity and calling behavior in wild mice

Background Animals rely on sound to mediate a myriad of daily activities, and anthropogenic noise is a pollutant that alters the natural soundscape within which they are active. As human infrastructure expands, broadband anthropogenic noise increases, which can affect behaviors of free-living nocturnal animals. Mice are nocturnal animals that produce ultrasonic calls as part of their behavioral repertoire. Methods We assessed effects of anthropogenic and natural noise on the behaviors of wild deer mice (Peromyscus maniculatus) and woodland jumping mice (Napaeozapus insignis), two species of mice that produce ultrasonic calls. We measured activity, foraging behavior at a foraging tray, and calling behavior to broadcasts of natural and anthropogenic noise, compared to a baseline with no broadcasting, at 25 focal areas in the Southern Appalachian Mountain Range of North Carolina, USA. Results Deer mice exposed to anthropogenic noise spent less time in focal areas with broadcasted anthropogenic noise. Mice took longer to begin foraging in the presence of anthropogenic noise, they spent less time at the foraging tray, and left fewer husks but consumed the same number of seeds as mice exposed to natural noise. Deer mice were less likely than woodland jumping mice to be the first to enter the focal area and approach food when in the presence of anthropogenic noise. Both species produced few ultrasonic calls in the presence of broadcasted natural and anthropogenic noise compared to their baseline level of calling. We present the first calls recorded from woodland jumping mice. Conclusion Anthropogenic noise affects activity, foraging behavior, and calling behavior of nocturnal mice. Natural noise also affects the calling behavior of mice. Mouse species respond differently to anthropogenic noise, with deer mice appearing more sensitive to anthropogenic noise than woodland jumping mice. Responses to noise could have important effects on the ecology of mice and these two species respond differently. Species differences should be considered when mitigating the effects of noise in conservation ecology.

Comparison of impacts of oil infrastructure on grassland songbirds between landscape-scale noise addition and noise removal experiments

As anthropogenic noise becomes increasingly widespread in natural habitats, noise addition and removal studies have become an important and commonly used method to assess the effects of noise on wildlife. Despite their wide implementation, it is difficult to determine whether the results from these studies translate to natural environments. Oil extraction operations provide a unique opportunity to conduct large-scale natural and experimental noise studies because they emit high-amplitude noise and are subject to interruptions or cessation of operations. We conducted a 6-year study aimed at identifying the effects of different types of oil infrastructure and noise on four species of grassland songbirds in the Canadian prairies. We measured abundance and nesting success of our focal species and compared between a large-scale playback experiment (i.e., noise addition) and in response to real infrastructure (i.e., noise removal). We also conducted an in-depth species analysis of reproductive output, stress, and population structure of an endangered species, Chestnut-collared Longspur ( Calcarius ornatus ). This multi-species and species-specific approach enabled us to assess noise impacts that may vary among species and for different life-history traits. Overall, our comparative study suggests that noise addition experiments can result in different conclusions regarding the ecological impacts of noise generated from in situ field studies in sites with associated disturbances. While noise clearly impacts birds in many ways, when layered with the many other ecosystem changes associated with real-world oil extraction activity, the effects of noise may be less prominent. As such, caution must be taken when applying results of noise experiments to conservation and management plans and regulations, as effects of noise predicted by lab and field noise addition experiments may simply not be realized under real-world conditions.

Efficient Generation of Paired Single-Cell Multiomics Profiles by Deep Learning.

Recent advances in single-cell sequencing technology have made it possible to measure multiple paired omics simultaneously in a single cell such as cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq) and single-nucleus chromatin accessibility and mRNA expression sequencing (SNARE-seq). However, the widespread application of these single-cell multiomics profiling technologies has been limited by their experimental complexity, noise in nature, and high cost. In addition, single-omics sequencing technologies have generated tremendous and high-quality single-cell datasets but have yet to be fully utilized. Here, single-cell multiomics generation (scMOG), a deep learning-based framework to generate single-cell assay for transposase-accessible chromatin (ATAC)data in silico is developed from experimentally available single-cell RNA-seq measurements and vice versa. The results demonstrate that scMOG can accurately perform cross-omics generation between RNA and ATAC, and generate paired multiomics data with biological meanings when one omics is experimentally unavailable and out of training datasets. The generated ATAC, either alone or in combination with measured RNA, exhibits equivalent or superior performance to that of the experimentally measured counterparts throughout multiple downstream analyses. scMOG is also applied to human lymphoma data, which proves to be more effective in identifying tumor samples than the experimentally measured ATAC data. Finally, the performance of scMOG is investigated in other omics such as proteomics and it still shows robust performance on surface protein generation.

An Efficient Approach to Manage Natural Noises in Recommender Systems

Recommender systems search the underlying preferences of users according to their historical ratings and recommend a list of items that may be of interest to them. Rating information plays an important role in revealing the true tastes of users. However, previous research indicates that natural noises may exist in the historical ratings and mislead the recommendation results. To deal with natural noises, different methods have been proposed, such as directly removing noises, correcting noise by re-predicting, or using additional information. However, these methods introduce some new problems, such as data sparsity and introducing new sources of noise. To address the problems, we present a new approach to managing natural noises in recommendation systems. Firstly, we provide the detection criteria for natural noises based on the classifications of users and items. After the noises are detected, we correct them with threshold values weighted by probabilities. Experimental results show that the proposed method can effectively correct natural noise and greatly improve the quality of recommendations.

Image encryption algorithm based on 1D-SFACF with cross-cyclic shift and adaptive diffusion

In this paper, a new one-dimensional chaotic system is proposed, which is called one-dimensional sine-fractional-adjusted-cosine-fractional (1D-SFACF). The dynamic behavior of the 1D-SFACF is analyzed by LE, bifurcation graph, trajectory, and cobweb plot. Compared to most one-dimensional chaotic systems, the 1D-SFACF has good cryptographic properties, such as larger parameter space and better chaotic behavior. Based on the excellent expressiveness of the 1D-SFACF, we design a secure image encryption algorithm using 1D-SFACF, called the image encryption algorithm based on 1D-SFACF (SFACF-IE). SFACF-IE is divided into four steps. First, the control parameters and initial values of 1D-SFACF are generated through a natural noise using a hash function. Secondly, an adaptive diffusion strategy is proposed. The starting position of the plaintext diffusion is adaptively changed according to the characteristics of the plaintext. Then, use a cross-cyclic shift to scramble. Finally, a hash function is used to generate a feedback key, the scrambled image is the input of the hash function, and the ciphertext is obtained by diffusing again. Experimental evaluations show that the algorithm is resistant to common attacks.

PKAID-Net: Prior Knowledge Aware Iterative Denoising Neural Network for Photon Counting Detector CT.

An important feature enabled by Photon-Counting Detector (PCD) CT is the simultaneous acquisition of multi-energy data, which can produce virtual monoenergetic images (VMIs) at a high spatial resolution. However, noise levels observed in the high-resolution (HR) VMIs are markedly increased. Recent work involving deep learning methods has shown great potential in CT image denoising. Many CNN applications involve training using spatially co-registered low- and high-dose CT images featuring high and low image noise, respectively. However, this is implausible in routine clinical practice. Further, typical denoising methods treat each VMI energy level independently, without consideration of the valuable information in the spectral domain. To overcome these obstacles, we propose a prior knowledge-aware iterative denoising neural network (PKAID-Net). PKAID-Net offers two major benefits: first, it utilizes spectral information by including a lower-noise VMI as a prior input; and second, it iteratively constructs refined datasets for neural network training to improve the denoising performance. This study includes 10 patient coronary CT angiography (CTA) exams acquired on a clinical HR PCD-CT (NAEOTOM Alpha, Siemens Healthineers). The HR VMIs were reconstructed at 50 and 70 keV, using a sharp kernel (Bv68) and thin (0.6 mm, 0.3 mm increment) slice thickness. Results showed that the PKAID-Net provided a noise reduction of 96% and 70% relative to FBP and iterative reconstruction, respectively while maintaining spatial and spectral fidelity and a natural noise texture. These results demonstrate the noise reduction capacity of PKAID-Net as applied to cutting-edge PCD-CT data to enable HR, multi-energy cardiac CT imaging.

The Social Determinants of Loneliness During COVID-19: Personal, Community, and Societal Predictors and Implications for Treatment

AbstractObjectiveThe COVID-19 pandemic dramatically altered social determinants of health including work, education, social connections, movement, and perceived control; and loneliness was commonly experienced. This longitudinal study examined how social determinants at the personal (micro), community (meso), and societal (macro) levels predicted loneliness during the pandemic.MethodsParticipants were 2056 Australian adults surveyed up to three times over 18 months in 2020 and 2021. Multi-level mixed-effect regressions were conducted predicting loneliness from social determinants at baseline and two follow-ups.ResultsLoneliness was associated with numerous micro determinants: male gender, lifetime diagnosis of a mental health disorder, experience of recent stressful event(s), low income, living alone or couples with children, living in housing with low natural light, noise, and major building defects. Lower resilience and perceived control over health and life were also associated with greater loneliness. At the meso level, reduced engagement with social groups, living in inner regional areas, and living in neighbourhoods with low levels of belongingness and collective resilience was associated with increased loneliness. At the macro level, increased loneliness was associated with State/Territory of residence.ConclusionsTherapeutic initiatives must go beyond psychological intervention, and must recognise the social determinants of loneliness at the meso and macro levels.

Acoustical and ventilation performance of school building near airport area in relation to Indoor Environmental Quality (IEQ)

Buildings that prioritize natural ventilation have problems dealing with airborne noise. The Indoor Environmental Quality (IEQ) disturbance will be considered in the school design located around the airport. As a result, the disorder can cause Sick Building Syndrome (SBS). Thus, the aspect of natural ventilation and noise (acoustics) is a problem that must be compromised. This paper will report the results of measuring the IEQ parameters of ventilation and acoustic, indicated by changes in the ratio of window openings in an elementary school located closest to the airplane runway. The results show that the top-hung type opening with a ratio of the half from the maximum opening produces airflow and noise that still does not meet the IEQ requirements. Hence the aspect of temperature and humidity and CO2 levels still meets the IEQ requirements.

Sampling and noise filtering methods for recommender systems: A literature review

In the era of online business, many e-commerce sites have evolved which recommend items according to one’s needs and interests. Plenty of data is available to be processed to make the recommender systems work effectively and efficiently. But, processing the entire dataset is a cumbersome process. So, there is a need to select a part of the data to be processed easily. Sampling is a way to select a subset of the entire dataset which contains all the attributes that the database can represent. It is important to understand which type of sampling method is suitable for a particular application in recommender systems. Thus, there is a need to study various sampling methods previously used in recommender systems. Also, before processing the data, we need to clean it up as it contains a certain amount of noise. This noise is described as either malicious noise or natural noise. Malicious noise is implicitly inserted in the system to alter the behavior of the system. This type of noise is termed as shilling attack. Natural noise enters the systems unknowingly due to the reluctance of users in giving proper ratings to the items. So, there is a need to filter these types of noise before making the data suitable for processing. In this paper, we have provided a review of 80 papers including both sampling and noise filtering methods. This is the first paper to the best of our knowledge, combining a literature review of both sampling and noise filtering methods.

The Linkages Between Urban Green Spaces and Residents’ Wellbeing: A Conceptual Framework

Sustainable development has been one of the biggest challenges the world has ever faced in recent decades. The term sustainable development is a dynamic process which brings together global and local concerns while linking social, economic, and environmental issues in order to resolve the needs of both current and future generations. It is therefore important to achieve sustainable development as urban residents' wellbeing, meaning their health, sense of security, freedom and social relations with others are heavily dependent upon it. In the modern era, sustainable development has become even more challenging because the nature of urban development which is often intertwined with urban issues such as urban sprawl, immigration, city planning. Similarly in Malaysia, the rapid urbanisation, transformation and expansion of urban cities has pushed the country towards the need for urban sustainability. Nonetheless, the destruction of natural green spaces, noise and overcrowding is one of the negative implications in compact cities that threatens urban residents’ wellbeing. This is worrisome because the urban green spaces are essential in promoting social interaction, health and wellbeing. Therefore, urban green spaces are among the most important sustainable urban development elements to integrate in residential development initiatives to achieve residents’ wellbeing. This paper provides insight into the function of spatial values of these urban green spaces in explaining residents' wellbeing. In this paper, there are three dimensions namely urban green spatial characteristics, urban green spatial value, and residents' wellbeing. This paper posits that favourable urban green spatial characteristics improves the perceptions of urban green spatial values, and consequently promotes residents' sense of wellbeing.

Deep learning-based reconstruction can improve the image quality of low radiation dose head CT.

To evaluate the image quality of deep learning-based reconstruction (DLR), model-based (MBIR), and hybrid iterative reconstruction (HIR) algorithms for lower-dose (LD) unenhanced head CT and compare it with those of standard-dose (STD) HIR images. This retrospective study included 114 patients who underwent unenhanced head CT using the STD (n = 57) or LD (n = 57) protocol on a 320-row CT. STD images were reconstructed with HIR; LD images were reconstructed with HIR (LD-HIR), MBIR (LD-MBIR), and DLR (LD-DLR). The image noise, gray and white matter (GM-WM) contrast, and contrast-to-noise ratio (CNR) at the basal ganglia and posterior fossa levels were quantified. The noise magnitude, noise texture, GM-WM contrast, image sharpness, streak artifact, and subjective acceptability were independently scored by three radiologists (1 = worst, 5 = best). The lesion conspicuity of LD-HIR, LD-MBIR, and LD-DLR was ranked through side-by-side assessments (1 = worst, 3 = best). Reconstruction times of three algorithms were measured. The effective dose of LD was 25% lower than that of STD. Lower image noise, higher GM-WM contrast, and higher CNR were observed in LD-DLR and LD-MBIR than those in STD (all, p ≤ 0.035). Compared with STD, the noise texture, image sharpness, and subjective acceptability were inferior for LD-MBIR and superior for LD-DLR (all, p < 0.001). The lesion conspicuity of LD-DLR (2.9 ± 0.2) was higher than that of HIR (1.2 ± 0.3) and MBIR (1.8 ± 0.4) (all, p < 0.001). Reconstruction times of HIR, MBIR, and DLR were 11 ± 1, 319 ± 17, and 24 ± 1s, respectively. DLR can enhance the image quality of head CT while preserving low radiation dose level and short reconstruction time. •For unenhanced head CT, DLR reduced the image noise and improved the GM-WM contrast and lesion delineation without sacrificing the natural noise texture and image sharpness relative to HIR. •The subjective and objective image quality of DLR was better than that of HIR even at 25% reduced dose without considerably increasing the image reconstruction times (24s vs. 11s). •Despite the strong noise reduction and improved GM-WM contrast performance, MBIR degraded the noise texture, sharpness, and subjective acceptance with prolonged reconstruction times relative to HIR, potentially hampering its feasibility.

A novel method for processing noisy magnetotelluric data based on independence of signal sources and continuity of response functions

Various schemes for processing magnetotelluric (MT) data have been reported aiming at suppressing the strong effect of artificial electromagnetic noises, especially coherent noise that is correlated between electric and magnetic time series. Many of the recent denoising schemes are based on decomposing MT data into the responses of the natural signals and noises. However, a steady differentiation of the natural signals from noises independent of any empirical choice of parameter setting is critically important. In addition, improper subtraction of values from the separated signals can lead to losing useful values of the natural signals or missing noise-affected values, which may result in failure in deriving the true MT responses. We propose a novel data-processing method that applies frequency-domain independent component analysis (FDICA) to both the local MT data and the reference magnetic data. Among the separated signals, the proposed method can quantitatively distinguish the natural signals from the noise-affected components by calculating the ratio of the cross-power spectrum with the reference data to the auto-power spectrum for each component. When determining which values to subtract from the separated signals, we introduce an evaluation index with respect to two characteristics of the MT response function: stationarity in the time domain and smoothness in the frequency domain. We conducted two types of experiments: with MT time series severely contaminated by synthetic coherent noises and with MT field data interfered with direct-current (DC) railways. Consequently, we confirmed the noise-suppression performance superiority of the proposed method over the conventional methods of MT data processing.

Testing Velocity-field Lensing on IllustrisTNG Galaxies

Weak gravitational lensing shear could be measured far more precisely if information about unlensed attributes of source galaxies were available. Disk galaxy velocity fields supply such information, at least in principle, with idealized models predicting orders of magnitude more Fisher information when velocity-field observations are used to complement images. To test the level at which realistic features of disk galaxies (warps, bars, spiral arms, and other substructure) inject noise or bias into such shear measurements, we fit an idealized disk model, including shear, to unsheared galaxies in the Illustris TNG100 simulation. The inferred shear thus indicates the extent to which unmodeled galaxy features inject noise and bias. We find that γ +, the component of shear parallel to the galaxy’s first principal axis, is highly biased and noisy because disks violate the assumption of face-on circularity, displaying a range of intrinsic axis ratios (0.85 ± 0.11). The other shear component, γ ×, shows little bias and is well-described by a double Gaussian distribution with central core scatter σ core ≈ 0.03, with low-amplitude, broad wings. This is the first measurement of the natural noise floor in the proposed velocity-field lensing technique. We conclude that the technique will achieve impressive precision gains for measurements of γ ×, but little gain for measurements of γ +.

Blown away? Wind speed and foraging success in an acoustic predator

Foraging animals must contend with fluctuating environmental variables that affect foraging success, including conditions like wind noise, which could diminish the usefulness of particular sensory modes. Although the documented impact of anthropogenic noise on animal behavior has become clear, there is limited research on natural noise and its potential influence on mammalian behavior. We investigated foraging behavior in the myrmecophagous bat-eared fox (Otocyon megalotis), a species known to rely predominantly on hearing for prey detection. For a year, we monitored the foraging behavior of 18 bat-eared foxes from a habituated population in South Africa, amidst varying wind speeds (0–15.5 km/h). In contrast to expectations, foraging rates did not generally decline with increasing wind speed, except for foraging rate outside termite patches in fall. Furthermore, wind speed had little correlation with time spent in patches. In winter, however, we observed an increase in foraging rate with increasing wind speed both within and outside patches. At the observed wind speeds, these acoustically driven insectivores continue to forage effectively despite potentially distracting or masking noises. With anthropogenic noise producing sound across a broader frequency range, it is important to examine the responses of these canids to artificial sources of acoustic disturbance as well.

Hybrid loss and domain transform based seismic image blind denoising

Noise removal is a critical stage in the processing of seismic data. Recent data-driven approaches require pre-assumptions about the distribution of noise for training the model, but in the actual denoising work, the natural noise does not obey the assumed distribution and the ability of the network to deal with this kind of noise is limited. Therefore, this paper proposes an adversarial learning method, FDT, which based on hybrid loss and domain transform. It treats noisy seismic data and clean seismic data as two different domains, and performs adversarial training by cycling between the two domains. Then the network will automatically learn a mapping that converts noisy seismic data into clean seismic data, which can achieve promising denoising result despite the noisy distribution is unknown. Moreover, single pixel-level loss similar to L1 and L2 is not adequate to characterize the texture and structure of the seismic data, in this paper, we combined SSIM(Structure Similarity Index Measure) and L1 together as a reconstruction loss to ensure that seismic data remains the same geological and tectonic information after denoising. Experimental results demonstrate that the method in this paper can effectively remove most types of noise from seismic data and achieve advanced results on both synthetic and real datasets. And it outputs a PSNR(Peak Signal-to-Noise Ratio) value of 27.97 with a standard deviation of 0.2 for large scale noise, maintaining a highly reliable denoising performance.

Broadband potential optimisation of a full scale acoustic metawindow performance.

Noise control and airflow in duct-like systems are among some of the most interesting applications to conjugate AMMs innovation and sustainability. Specifically applied to the built environment, they opened up a new field of research supporting indoor wellbeing, sanitised environments, and public activities. Previous research conducted by the authors has proved AMM based window to be a resourceful way to address both natural ventilation and reduce the incoming noise propagation; however, the effective spectral range did not cover lower frequencies (50-350 Hz). For this reason, in the presented paper, implementation in the AMM unit geometry was performed over a full-scale acoustic metawindow (AMW). The resonating volume has been enhanced (by 200% of the original one) and coupled with a set of lateral flanks. Numerical analysis through FEM proved that on a range of opening ratio from 3 to 33%, the TL related to the window is improved overall of the 70% on the frequency range from 50 to 350 Hz. Such results encourage the use of new AMMs ergonomic windows in place of standard ones to achieve both natural ventilation and noise attenuation from 50 to 5k Hz, being resourceful for domestic, sanitary, and public applications.

Autoencoder-Based Recommender System Exploiting Natural Noise Removal

Collaborative filtering (CF) is a widely used technique in recommender systems by automatically predicting the user’s latent interests based on many users’ historical rating data. To improve the performance of the CF-based recommender systems, users’ rating data should be pre-processed to avoid noise and enhance data reliability. Many researchers studied anomaly detection to remove malicious noise caused by shilling attacks, but anomalies can still exist in non-attacked real user data, which is called natural noise, as the ratings of users can be impacted by unpredictable factors such as other users’ ratings and anchoring bias. In this paper, we propose an autoencoder-based recommendation system for exploiting the ability of both anomaly detection and CF. The proposed system detects the natural noise in the rating data based on the reconstruction errors after training. By removing the detected natural noise, CF can predict the unrated ratings with noise-free data. Our experiments show that the proposed model showed better performance than the traditional method by reducing the error by up to 5% compared to the method that does not consider natural noise detection and reducing the error by up to 4% compared to the conventional rating classification based natural noise detection methods.

Comparison of spectral subtraction noise reduction algorithms

Noise in media is any undesirable signal that masks relevant information content. The addition of noise to real-world data in any context is practically inevitable. Noise reduction algorithms in the past have addressed the problem but lacked adaptability to various real-world applications while also being time and resource extensive. Spectral subtraction provides a hybrid approach to noise reduction that incorporates versatility and efficient resource usage. This research tested the performance of two spectral subtraction noise reduction algorithms (stationary and non-stationary) across five categories of real-world noise (speech only, speech with natural noise, music, animal sounds, and noise only). The research question under study was how stationary and non-stationary spectral subtraction algorithms differ in their noise reduction performance when subjected to the various categories of noise. The testing was done based on normalized cross-correlation, which is the similarity between the noise-reduced audio and the original recording in each case. Non-stationary spectral subtraction performed better in samples where human speech was the target: speech only and speech with natural noise. Stationary spectral subtraction performed better when denoising music and animal sounds. This anomaly in performance between the two algorithms was only noted in categories with no human speech. These results exemplify the performance and versatility of different spectral subtraction algorithms. The category-specific results can be used to employ specific spectral subtraction algorithms for specific tasks for optimum performance.

Enhancement of the Seismic Data Resolution Through Q-Compensated Denoising Based on Dictionary Learning

Seismic wave acquisition is usually disturbed by natural noise and instrument noise. As the seismic wave propagates, the filtering effect of the Earth and its various layers will result in energy attenuation and velocity dispersion; these phenomena weaken the seismic time series amplitudes and distort the seismic phase data. In traditional processing methods, noise removal precedes the data compensation process, which attempts to retrieve information that was originally lost due to signal attenuation. If denoising is performed, weak seismic signals are often removed, resulting in the loss of useful signal data that cannot be recovered. Here, we propose a sparse regularization strategy based on dictionary learning. This inversion method performs the denoising and data compensation tasks in parallel. By applying this method to both synthetic and real datasets, we demonstrate that this technique effectively compensates and denoises the seismic data and improves both the data resolution and the signal-to-noise ratio of the seismic records of interest.