Masking Functions Research Articles

Approximate Triple Modular Redundancy: A Survey

In recent years, approximate computing (AC) has attracted attention owing to its tradeoff between the exactness of computations and performance gains. AC has also been probed for the technique of Triple modular redundancy (TMR). TMR is a well-known fault masking methodology, with associated overheads, widely used in systems of different nature and at different levels. E.g.: layout-level, gate-level, HW-module level, software. At hardware level, through exploitation of AC the 200% area overhead problem due to triplication of the original modules in TMR can be reduced. By approximating the modules of TMR while ensuring that at least two of the approximate modules do not differ from the original module for every input vector, the facilitation of fault masking can lead to overhead reduction. Hence, approximate TMR (ATMR) aims to achieve cost-effective reliability. Nevertheless, due to the extensive search space, computational complexity, and principal fault masking function of ATMR, designing an ATMR is a challenging task. An ATMR technique must be scalable so that it can be easily adopted by circuits having large number of inputs and the extraction of ATMR modules remains computationally inexpensive. Compared with TMR, due to the inclusion of approximations, ATMR is more vulnerable to errors, and hence, the design technique must ensure awareness of input-criticality. To the best of the authors' knowledge, none of the existing survey articles on AC has reported on ATMR. Therefore, in this work, ATMR design techniques are thoroughly surveyed and qualitatively compared. Moreover, design considerations and challenges for designing ATMR are discussed.

Mask family proteins ANKHD1 and ANKRD17 regulate YAP nuclear import and stability.

Mask family proteins were discovered in Drosophila to promote the activity of the transcriptional coactivator Yorkie (Yki), the sole fly homolog of mammalian YAP (YAP1) and TAZ (WWTR1). The molecular function of Mask, or its mammalian homologs Mask1 (ANKHD1) and Mask2 (ANKRD17), remains unclear. Mask family proteins contain two ankyrin repeat domains that bind Yki/YAP as well as a conserved nuclear localisation sequence (NLS) and nuclear export sequence (NES), suggesting a role in nucleo-cytoplasmic transport. Here we show that Mask acts to promote nuclear import of Yki, and that addition of an ectopic NLS to Yki is sufficient to bypass the requirement for Mask in Yki-driven tissue growth. Mammalian Mask1/2 proteins also promote nuclear import of YAP, as well as stabilising YAP and driving formation of liquid droplets. Mask1/2 and YAP normally colocalise in a granular fashion in both nucleus and cytoplasm, and are co-regulated during mechanotransduction.

Unsupervised Low Latency Speech Enhancement With RT-GCC-NMF

In this paper, we present RT-GCC-NMF: a real-time (RT), two-channel blind speech enhancement algorithm that combines the non-negative matrix factorization (NMF) dictionary learning algorithm with the generalized cross-correlation (GCC) spatial localization method. Using a pre-learned universal NMF dictionary, RT-GCC-NMF operates in a frame-by-frame fashion by associating individual dictionary atoms to target speech or background interference based on their estimated time-delay of arrivals (TDOA). We evaluate RT-GCC-NMF on two-channel mixtures of speech and real-world noise from the Signal Separation and Evaluation Campaign (SiSEC). We demonstrate that this approach generalizes to new speakers, acoustic environments, and recording setups from very little training data, and outperforms all but one of the algorithms from the SiSEC challenge in terms of overall Perceptual Evaluation methods for Audio Source Separation (PEASS) scores and compares favourably to the ideal binary mask baseline. Over a wide range of input SNRs, we show that this approach simultaneously improves the PEASS and signal to noise ratio (SNR)-based Blind Source Separation (BSS) Eval objective quality metrics as well as the short-time objective intelligibility (STOI) and extended STOI (ESTOI) objective speech intelligibility metrics. A flexible, soft masking function in the space of NMF activation coefficients offers real-time control of the trade-off between interference suppression and target speaker fidelity. Finally, we use an asymmetric short-time Fourier transform (STFT) to reduce the inherent algorithmic latency of RT-GCC-NMF from 64 ms to 2 ms with no loss in performance. We demonstrate that latencies within the tolerable range for hearing aids are possible on current hardware platforms.

Detection and intensity discrimination of a sinusoid with and without masker uncertainty

The effect of stimulus-based uncertainty on the shape of the masking function at circa-threshold pedestal levels was investigated. Of particular interest was how uncertainty regarding properties of the masker affects the magnitude of “negative masking” [Raab et al., J. Acoust. Soc. Am. 35, 1053 (1963)] that is obtained under different stimulus configurations. Intensity discrimination thresholds for gated, 100-ms, 1000-Hz sinusoids were measured at three pedestal levels: -9, 0, and 9 dB re: absolute threshold. A two-interval, two-alternative forced-choice procedure was used. Under the reference condition, thresholds were measured in quiet. In comparison conditions, thresholds were measured in the presence of one of two masker types: (1) a notched-noise masker or (2) a random-frequency, multicomponent masker. In the multicomponent masker condition, uncertainty was imposed by roving the frequency components that comprised the masker from interval-to-interval over a block of trials. The data under each condition were fit with psychometric functions and slope estimates were obtained. Results are discussed with respect to the factors that govern, and possible mechanisms underlying, negative masking. Specifically, two possible explanations for the “dip” in the masking function are examined: nonlinear transduction [Hanna et al., J. Acoust. Soc. Am. 80, 1335 (1986)] and observer uncertainty.

2.2.2.2](2,7)‐1‐Bromonaphthalenophane from a Desymmetrized Building Block Bearing Electrophilic and Masked Nucleophilic Functionalities

AbstractIn search of 2,7‐ethylene‐bridged naphthalenophanes with desymmetrized naphthalene cores as inherently chiral cyclophanes, nucleophilic substitution of 1‐bromo‐7‐(bromomethyl)‐2‐[(trimethylsilyl)methyl]naphthalene, a desymmetrized building block bearing an electrophilic group (CH2Br) and a masked nucleophilic functionality (CH2TMS) which can be activated by fluoride anion, was examined. As a result, in contrast to the case of parent naphthalenophanes wherein the smallest [2.2]naphthalenophane was obtained as the major product, only [2.2.2.2](2,7)‐1‐bromonaphthalenophane was obtained albeit in low yields, whereas the corresponding [2.2]‐ or [2.2.2]naphthalenophanes were not obtained. Though the [2.2.2.2]‐1‐bromonaphthalenophane can adopt four idealized geometries of different symmetry, among which three are inherently chiral, theoretical calculations predict that three conformers have almost equal energy and may equilibrate in solution. The X‐ray crystallographic study shows that it adopts a C2 conformation with anti,anti,anti orientation of the C−Br bonds at least as a major component in crystal.

A versatile de novo synthesis of legionaminic acid and 4-epi-legionaminic acid starting from d-serine

Legionaminic acid and 4-epi-legionaminic acid are 5,7-diacetamido nonulosonic acids and are assumed to play a crucial role in the virulence of Legionella pneumophila, the causative agent of Legionnaires’ disease. Moreover, they are ideal target motifs for the development of vaccines and pathogen detection. Herein, we present a versatile de novo synthesis of legionaminic acid and 4-epi-legionaminic acid. Starting from simple d-serine, the C9-backbone is built up by two CC-bond formation reactions. First, the protected d-serine motif is elongated utilizing a highly stereoselective nitroaldol reaction to give a C6-precursor of desired d-rhamno configuration. Second, an indium-mediated allylation is employed to further elongate the carbon backbone and introduce a masked α-keto acid function.

Development of Co-processed Paracetamol with Hydroxypropyl Methylcellulose (HPMC) and Maltodextrin by Wet Granulation Process

The purpose of this work is to improve one of the important physicochemical properties of an active pharmaceutical ingredient (API), as is the water solubility of paracetamol. To improve the physicochemical properties of this API, two pharmaceutical excipients such as the HPMC and maltodextrin were used, which help to improve the solubility, and this helps to the manufacturing process of a pharmaceutical product. Different granule formulations were manufactured by applying a matrix design of experiment where the wet granulation process was performed, combining Paracetamol with the excipients to obtain a uniform particle size and subsequently evaluate the properties of interest. The solubility was evaluated using a method (Mexican pharmacopoeia - FEUM) based on UV / VIS method, performing the calibration curve only for the API to evaluate the granule and calculate the percentage of solubility of these. Favorable results were obtained for two of the seven granule formulations manufactured, the mixture of granule F and G: 25 g of paracetamol, 1.5; 1.75 g of HPMC and 23.5 g; 23.25 of the maltodextrin has a solubility of 104.17% and 101.48% of the G, which shows that the process by wet granulation can improve its solubility. This type of co-processed granule also fulfills the function of masking the bitter taste of paracetamol in one oral pharmaceutical form, as in the case of a syrup and it could be an advantage in the market. The flavor was evaluated by a panel of 20 people and the taste of the syrups that were prepared with the granule with better solubility was compared with the syrups containing only the API dissolved. It is shown that the granules F have improvement in the solubility of paracetamol and can mask the unpleasant (bitter) taste of the active ingredient.

Investigation of α clustering with knockout reactions

Our goal is to reveal how the $\alpha$ cluster amplitude is probed through $\alpha$ knockout reactions depending on reaction conditions, e.g., the incident energy. We consider $^{20}$Ne($p$,$p\alpha$)$^{16}$O and $^{120}$Sn($p$,$p\alpha$)$^{116}$Cd at 100-400 MeV within the distorted wave impulse approximation (DWIA) framework. We introduce a masking function which shows how the reaction amplitude in the nuclear interior is suppressed and defines the probed region of the $\alpha$ cluster wave function. It is clearly shown by means of the masking function that the $\alpha$ knockout reaction probes the $\alpha$ cluster amplitude in the nuclear surface region, which is the direct measure of well-developed $\alpha$ cluster states. A simplified form of the masking function is introduced and the incident energy dependence of the masking effect is investigated. $\alpha$ knockout reaction can probe the $\alpha$ cluster amplitude in the nuclear surface region by choosing proper kinematics owing to the masking effect originated from absorptions of distorting potentials, and is a suitable method to investigate how $\alpha$ cluster states are spatially developed.

A new Landsat 8 cloud discrimination algorithm using thresholding tests

ABSTRACTIn this work, we propose a Cloud Discrimination Algorithm for Landsat 8 (CDAL8) to improve a high-frequency automatic land change detection system developed at the National Institute of Advanced Industrial Science and Technology (AIST), Japan for large-scale satellite image analysis. Although the land change detection system can process several kinds of satellite remote sensing data, improvements are needed to enable practical applications using Landsat 8 data. Cloud discrimination is a necessary pre-processing step for land cover change detection. Currently, most of the prediction errors on land change detection are caused by the false cloud discrimination results as a pre-processing step. Therefore, we introduce an improved cloud discrimination algorithm (CDAL8) in this study to improve the overall performance of our land change detection system. The algorithm was developed based on a Moderate Resolution Imaging Spectroradiometer (MODIS) cloud mask algorithm and Cloud and Aerosol Unbiased Decision Intellectual Algorithm (CLAUDIA). CDAL8 is distinct in that it switches judgment tests and their thresholds using a threshold brightness temperature and uses separate features in cloud judgment and clear-sky judgment. To evaluate the accuracy of the proposed algorithm, we compared it with the Automated Cloud-Cover Assessment algorithm (ACCA) and Function of Mask (Fmask) version 3.3 using US Geological Survey Landsat 8 cloud cover assessment validation data, which contain 96 cloud masks. Our proposed cloud discrimination algorithm (CDAL8) have promising results with an accuracy of 88.1%, which was greater than that of the ACCA (82.5%) and Fmask (84.6%). Furthermore, we also confirmed that the average accuracy of CDAL8 was approximately 91.2% when low solar elevation scenes were removed.

An automatic method for screening clouds and cloud shadows in optical satellite image time series in cloudy regions

Clouds and cloud shadows block land surface information in optical satellite images. Accurate detection of clouds and cloud shadows can help exclude these contaminated pixels in further applications. Existing cloud screening methods are challenged by cloudy regions where most of satellite images are contaminated by clouds. To solve this problem for landscapes where the typical frequency of cloud-free observations of a pixel is too small to use existing methods to mask clouds and shadows, this study presents a new Automatic Time-Series Analysis (ATSA) method to screen clouds and cloud shadows in multi-temporal optical images. ATSA has five main steps: (1) calculate cloud and shadow indices to highlight cloud and cloud shadow information; (2) obtain initial cloud mask by unsupervised classifiers; (3) refine initial cloud mask by analyzing time series of a cloud index; (4) predict the potential shadow mask using geometric relationships; and (5) refine the potential shadow mask by analyzing time series of a shadow index. Compared with existing methods, ATSA needs fewer predefined parameters, does not require a thermal infrared band, and is more suitable for areas with persistent clouds. The performance of ATSA was tested with Landsat-8 OLI images, Landsat-4 MSS images, and Sentinel-2 images in three sites. The results were compared with a popular method, Function of Mask (Fmask), which has been adopted by USGS to produce Landsat cloud masks. These tests show that ATSA and Fmask can get comparable cloud and shadow masks in some of the tested images. However, ATSA can consistently obtain high accuracy in all images, while Fmask has large omission or commission errors in some images. The quantitative accuracy was assessed using manual cloud masks of 15 images. The average cloud producer's accuracy of these 15 images is as high as 0.959 and the average shadow producer's accuracy reaches 0.901. Given that it can be applied to old satellite sensors and it is capable for cloudy regions, ATSA is a valuable supplement to the existing cloud screening methods.

Contrast thresholds reveal different visual masking functions in humans and praying mantises.

ABSTRACTRecently, we showed a novel property of the Hassenstein–Reichardt detector, namely that insect motion detection can be masked by ‘undetectable’ noise, i.e. visual noise presented at spatial frequencies at which coherently moving gratings do not elicit a response (Tarawneh et al., 2017). That study compared the responses of human and insect motion detectors using different ways of quantifying masking (contrast threshold in humans and masking tuning function in insects). In addition, some adjustments in experimental procedure, such as presenting the stimulus at a short viewing distance, were necessary to elicit a response in insects. These differences offer alternative explanations for the observed difference between human and insect responses to visual motion noise. Here, we report the results of new masking experiments in which we test whether differences in experimental paradigm and stimulus presentation between humans and insects can account for the undetectable noise effect reported earlier. We obtained contrast thresholds at two signal and two noise frequencies in both humans and praying mantises (Sphodromantis lineola), and compared contrast threshold differences when noise has the same versus different spatial frequency as the signal. Furthermore, we investigated whether differences in viewing geometry had any qualitative impact on the results. Consistent with our earlier finding, differences in contrast threshold show that visual noise masks much more effectively when presented at signal spatial frequency in humans (compared to a lower or higher spatial frequency), while in insects, noise is roughly equivalently effective when presented at either the signal spatial frequency or lower (compared to a higher spatial frequency). The characteristic difference between human and insect responses was unaffected by correcting for the stimulus distortion caused by short viewing distances in insects. These findings constitute stronger evidence that the undetectable noise effect reported earlier is a genuine difference between human and insect motion processing, and not an artefact caused by differences in experimental paradigms.

An optimized absorbing potential for ultrafast, strong-field problems

Theoretical treatments of strong-field physics have long relied on the numerical solution of the time-dependent Schrödinger equation. The most effective such treatments utilize a discrete spatial representation—a grid. Since most strong-field observables relate to the continuum portion of the wave function, the boundaries of the grid—which act as hard walls and thus cause reflection—can substantially impact the observables. Special care thus needs to be taken. While there exist a number of attempts to solve this problem—e.g., complex absorbing potentials and masking functions, exterior complex scaling, and coordinate scaling—none of them are completely satisfactory. The first of these is arguably the most popular, but it consumes a substantial fraction of the computing resources in any given calculation. Worse, this fraction grows with the dimensionality of the problem. In addition, no systematic way to design such a potential has been used in the strong-field community. In this work, we address these issues and find a much better solution. By comparing with previous widely used absorbing potentials, we find a factor of 3–4 reduction in the absorption range, given the same level of absorption over a specified energy interval.

Modern ways of improving the procedure for the formation and verification of digital signature

In the article, a comparative analysis of existing algorithms for the formation and verification of the digital signature has been carried out, namely, widespread algorithms and the criteria for comparison was the mathematical basics on which the algorithm is based and the size of the public key. Based on the analysis, it is established that there are schemes with the addition of messages and schemes with the restoration of the information message. The conducted studies made it possible to determine that the schemes with message recovery differ from the schemes with the addition that they do not completely hash messages, but instead use masking functions and redundancy of the message. It was determined that the most effective and optimal for further use is the Nimberg-Ruppel scheme, which is based on elliptical curves (ECNR). The article describes the theoretical basis for creating a system for ensuring the integrity, reliability and confidentiality of information based on digital signatures based on elliptical curves using the standard DSTU 4145-2002 with the possibility of restoring an information message based on ECNR.

A cloud shadow detection method combined with cloud height iteration and spectral analysis for Landsat 8 OLI data

Although enhanced over prior Landsat instruments, Landsat 8 OLI can obtain very high cloud detection precisions, but for the detection of cloud shadows, it still faces great challenges. Geometry-based cloud shadow detection methods are considered the most effective and are being improved constantly. The Function of Mask (Fmask) cloud shadow detection method is one of the most representative geometry-based methods that has been used for cloud shadow detection with Landsat 8 OLI. However, the Fmask method estimates cloud height employing fixed temperature rates, which are highly uncertain, and errors of large area cloud shadow detection can be caused by errors in estimations of cloud height. This article improves the geometry-based cloud shadow detection method for Landsat OLI from the following two aspects. (1) Cloud height no longer depends on the brightness temperature of the thermal infrared band but uses a possible dynamic range from 200 m to 12,000 m. In this case, cloud shadow is not a specific location but a possible range. Further analysis was carried out in the possible range based on the spectrum to determine cloud shadow location. This effectively avoids the cloud shadow leakage caused by the error in the height determination of a cloud. (2) Object-based and pixel spectral analyses are combined to detect cloud shadows, which can realize cloud shadow detection from two aspects of target scale and pixel scale. Based on the analysis of the spectral differences between the cloud shadow and typical ground objects, the best cloud shadow detection bands of Landsat 8 OLI were determined. The combined use of spectrum and shape can effectively improve the detection precision of cloud shadows produced by thin clouds. Several cloud shadow detection experiments were carried out, and the results were verified by the results of artificial recognition. The results of these experiments indicated that this method can identify cloud shadows in different regions with correct accuracy exceeding 80%, approximately 5% of the areas were wrongly identified, and approximately 10% of the cloud shadow areas were missing. The accuracy of this method is obviously higher than the recognition accuracy of Fmask, which has correct accuracy lower than 60%, and the missing recognition is approximately 40%.

Changing stimulation patterns can change the broadness of contralateral masking functions for bilateral cochlear implant users

Past studies have found that contralateral masking functions are sharper than ipsilateral masking functions for cochlear implant (CI) users. This could suggest that contralateral masking effects are only sensitive to the peak of the masker stimulation for this population. To determine if that is the case, this study investigated whether using broader stimulation patterns affects the broadness of the contralateral masking function. Contralateral masking functions were measured for six bilateral CI users using both a broad and narrow masker. Findings from this study revealed that the broad masker resulted in a broader contralateral masking function. This would suggest that stimulation outside of the peak of the masker affects contralateral masking functions for CI users.

Construction of Multivalent Homo- and Heterofunctional ABO Blood Group Glycoconjugates Using a Trifunctional Linker Strategy.

The design and synthesis of multivalent ligands displaying complex oligosaccharides is necessary for the development of therapeutics, diagnostics, and research tools. Here, we report an efficient conjugation strategy to prepare complex glycoconjugates with 4 copies of 1 or 2 separate glycan epitopes, providing 4-8 carbohydrate residues on a tetravalent poly(ethylene glycol) scaffold. This strategy provides complex glycoconjugates that approach the size of glycoproteins (15-18 kDa) while remaining well-defined. The synthetic strategy makes use of three orthogonal functional groups, including a reactive N-hydroxysuccinimide (NHS)-ester moiety on the linker to install the first carbohydrate epitope via reaction with an amine. A masked amine functionality on the linker is revealed after the removal of a fluorenylmethyloxycarbonyl (Fmoc)-protecting group, allowing the attachment to the NHS-activated poly(ethylene glycol) (PEG) scaffold. An azide group in the linker was then used to incorporate the second carbohydrate epitope via catalyzed alkyne-azide cycloaddition. Using a known tetravalent PEG scaffold (PDI, 1.025), we prepared homofunctional glycoconjugates that display four copies of lactose and the A-type II or the B-type II human blood group antigens. Using our trifunctional linker, we expanded this strategy to produce heterofunctional conjugates with four copies of two separate glycan epitopes. These heterofunctional conjugates included Neu5Ac, 3'-sialyllactose, or 6'-sialyllactose as a second antigen. Using an alternative strategy, we generated heterofunctional conjugates with three copies of the glycan epitope and one fluorescent group (on average) using a sequential dual-amine coupling strategy. These conjugation strategies should be easily generalized for conjugation of other complex glycans. We demonstrate that the glycan epitopes of heterofunctional conjugates engage and cluster target B-cell receptors and CD22 receptors on B cells, supporting the application of these reagents for investigating cellular response to carbohydrate antigens of the ABO blood group system.

Boundary-aware single fringe pattern demodulation

Optical interferometric techniques offer non-contact, high accuracy and full field measurements, which are very attractive in various research and application fields. Single fringe pattern processing is often needed when measuring fast phenomenon. However, several difficulties are encountered in phase retrieval, among which the discontinuity problem of the fringe pattern is challenging and requires attention due to the increasing complexity of manufactured pieces. In this paper, we propose a complete flowchart for discontinuous single fringe pattern processing, which uses segmentation as universal pre-processing for all discontinuous fringe pattern problems. Within the flowchart, we also propose a systematic way to introduce boundary-awareness into demodulation methods by a masking function to improve demodulation accuracy and a quality-guided scanning strategy with a novel composite quality map to improve demodulation robustness. To the best of our knowledge, this is the first time a complete solution is proposed for single and discontinuous fringe pattern processing. Three typical demodulation methods, the frequency-guided regularized phase tracker with quadratic phase matching, the windowed Fourier ridges, and the spiral phase quadrature transform, are used to demonstrate how demodulation methods can be made boundary-aware. The proposed methods are verified by successful demodulation results from both simulated and experimental fringe patterns.

Monitoring the dynamics of surface water fraction from MODIS time series in a Mediterranean environment

Detailed spatial information of changes in surface water extent is needed for water management and biodiversity conservation, particularly in drier parts of the globe where small, temporally-variant wetlands prevail. Although global surface water histories are now generated from 30m Landsat data, for many locations they contain large temporal gaps particularly for longer periods (>10years) due to revisit intervals and cloud cover. Daily Moderate Resolution Imaging Spectrometer (MODIS) imagery has potential to fill such gaps, but its relatively coarse spatial resolution may not detect small water bodies, which can be of great ecological importance. To address this problem, this study proposes and tests options for estimating the surface water fraction from MODIS 16-day 500m Bidirectional Reflectance Distribution Function (BRDF) corrected surface reflectance image composites. The spatial extent of two Landsat tiles over Spain were selected as test areas. We obtained a 500m reference dataset on surface water fraction by spatially aggregating 30m binary water masks obtained from the Landsat-derived C-version of Function of Mask (CFmask), which themselves were evaluated against high-resolution Google Earth imagery. Twelve regression tree models were developed with two approaches, Random Forest and Cubist, using spectral metrics derived from MODIS data and topographic parameters generated from a 30m spatial resolution digital elevation model. Results showed that accuracies were higher when we included annual summary statistics of the spectral metrics as predictor variables. Models trained on a single Landsat tile were ineffective in mapping surface water in the other tile, but global models trained with environmental conditions from both tiles can provide accurate results for both study areas. We achieved the highest accuracy with Cubist global model (R2=0.91, RMSE=11.05%, MAE=7.67%). Our method was not only effective for mapping permanent water fraction, but also in accurately capturing temporal fluctuations of surface water. Based on this good performance, we produced surface water fraction maps at 16-day interval for the 2000–2015 MODIS archive. Our approach is promising for monitoring surface water fraction at high frequency time intervals over much larger regions provided that training data are collected across the spatial domain for which the model will be applied.

A Deep Residual convolutional neural network for facial keypoint detection with missing labels

Keypoint detection is critical in image recognitions. Deep learning such as convolutional neural network (CNN) has recently demonstrated its tremendous success in detecting image keypoints over conventional image processing methodologies. The deep learning solutions, however, heavily rely on labeling target images for their reliability and accuracy. Unfortunately, most image datasets do not have all labels marked. To address this problem, this paper presents an effective and novel deep learning solution, Masked Loss Residual Convolutional Neural Network (ML-ResNet), to facial keypoint detection on the datasets that have missing target labels. The core of ML-ResNet is a masked loss objective function that ignores the error in predicting the missing target keypoints in the output layer of a CNN. To compensate for the loss induced by the masked loss objective function that likely results in overfitting, ML-ResNet is designed of a data augmentation strategy to increase the number of training data. The performance of ML-ResNet has been evaluated on the image dataset from Kaggle Facial Keypoints Detection competition, which consists of 7049 training images, but with only 2140 images that have full target keypoints labeled. In the experiments, ML-ResNet is compared to a pioneer literature CNN facial keypoint detection work. The experiment results clearly show that the proposed ML-ResNet is robust and advantageous in training CNNs on datasets with missing target values. ML-ResNet can improve the learning time by 30% during the training and the detection accuracy by eight times in facial keypoint detection.

Electrophysiological determination of backward masking function

Backward Masking (BM) functions have been shown to relate to age, lead toxicity, and in children with language disorders. These functions may be indicative of auditory processing deficits. This study investigated if Evoked Potentials (EP) could be utilized to obtain BM functions. The design of this study allowed observation of the early, middle, and late auditory evoked potentials. This process allowed us to observe the differential electrophysiological responses of evoked potentials during the BM effect. This study was randomized using four different stimulus conditions. With a long inter-stimulus interval and high sample rate simultaneous early, middle, and late potentials were obtained. The stimuli were a pure-tone alone, (1) 1000 Hz, (2) a masking noise alone, (3) the pure-tone followed by the masking noise, and (4) a control condition with no auditory stimulus. This approach, using a randomization allowed any adaptation or habituation to the stimuli to be equally distributed within each condition. Using arithmetic operations on the derived evoked potentials allows for localization of brain structures utilized in BM processing. Results indicated the determination of BM functions can be obtained objectively.