Marginal Distance Research Articles

Coastline target detection based on UAV hyperspectral remote sensing images

Timely and accurate monitoring of typical coastal targets using remote sensing technology is crucial for maintaining marine ecological stability. Hyperspectral target detection technology proves to be an effective tool in extracting various typical materials along the coastline. Traditional target detection methods using spectral domain information can effectively retain the intrinsic properties of the material. However, it is difficult to effectively recognize targets in homogeneous regions by using only spectral domain information, which may lead to insufficient utilization of spatial information. In this study, a detector based on signal-to-noise ratio fusion constrained energy minimization with low-rank sparse decomposition (SFLRSD) is proposed. This detector improves the separability of background and target by obtaining spatial domain information from hyperspectral images and fusing spectral domain information. First, total variation regularization and fractional Fourier transform are applied to process spatial and spectral domain information, respectively. The constrained energy minimization (CEM) detector is used to improve the separability between the target and background of the processed data. Then, the background and anomalies are represented as low-rank and sparse components, respectively, using low-rank sparse matrix factorization. This transforms the model solution into a covariance matrix problem, which is then solved using marginal distance difference (MDD) to isolate anomalous parts. Subsequently, the anomaly parts are fused with CEM detector results, weighted by their respective signal-to-noise ratios. This detection model leverages unified hyperspectral image features, enhancing spectral discreteness of anomalous targets and backgrounds. Finally, experiments on custom created hyperspectral dataset show that the proposed method outperforms other baseline methods in terms of visualization and quantitative performance. In this paper, we not only propose a new hyperspectral target detection method, but we also collect three typical marine litter of different materials by means of airborne hyperspectral remote sensing and construct four hyperspectral datasets in a real environment. All the simulation experiments in this paper are conducted in these four datasets.

Urban flood-bearing vulnerability evaluation based on the moment estimate weighting and improved gray target model.

Increasingly severe flooding seriously threatens urban safety. A scientific urban flood-bearing vulnerability assessment model is significant to improve urban risk management capacity. The gray target model (GTM) has advantages in urban flood-bearing vulnerability assessment. However, indicator correlation and single bull's-eye are commonly neglected, leading to defective evaluation results. By integrating the four base weights, an improved weighting method based on the moment estimate was proposed. Then, the marginal distance was used to quantify the indicator correlation, and the TOPSIS model was introduced to define the relative bull's-eye distance. Thus, an improved gray target evaluation method was established. Finally, an urban flood-bearing vulnerability evaluation model was presented based on the moment estimate weighting-improved GTM. In this study, Zhengzhou City, China, was taken as an example. The spatial and temporal changing characteristics of the flood-bearing vulnerability of Zhengzhou from 2006 to 2020 were investigated. The results show that: (1) On the temporal scale, the disaster-bearing vulnerability of Zhengzhou City showed an upward trend during the 15 years; (2) On the spatial scale, Guancheng District of Zhengzhou City had the relatively highest vulnerability to urban flooding. This study is expected to provide a scientific reference for urban flood risk management.

Risk-aware microgrid operation and participation in the day-ahead electricity market

This work examines the daily bidding problem of a grid-connected microgrid with locally deployed resources for electricity generation, storage and its own electricity demand. Trading electricity in energy markets may offer economic incentives but exposes the microgrid to financial risk caused by market commitments. Hence, a multi-objective, two-stage stochastic mixed integer linear programming (MILP) model is formulated, extending prior work of a risk-neutral microgrid bidding approach. The multi-objective model minimises both expected total cost of day-ahead microgrid operations and financial risk from bidding measured by conditional value-at-risk (CVaR). Bidding curves derived as first stage decisions are always feasible under present market rules – including a limitation on the number of break points per submitted curve – while being near optimal for the microgrid’s day-ahead recourse schedule. The proposed optimisation model is embedded in a variant of the ɛ-constrained method to generate bidding curve candidates with different trade-offs between the two conflicting objectives. Moreover, scenario reduction is used to compromise accuracy of the uncertainty set for better computational performance. Particularly, the marginal relative probability distance between initial and reduced scenario set is suggested to make a decision on the extent of scenario reduction. The proposed solution procedure is tested in a computational study to demonstrate its applicability to generate optimal microgrid bidding curve candidates with different emphasis between total cost and CVaR in reasonable computational time.

Look What We Have Here: Exploring Brand-Related Sport Consumer Twitter Conversation Topics

Abstract: As sport organizations leverage social media as a critical component of marketing strategy, tools for exploring the large volume of sport consumer social media conversations are vital. This scholarship demonstrates the value of unsupervised latent Dirichlet allocation (LDA) as a tool for exploring consumers' digital conversations. Specifically, unsupervised LDA was applied to derive latent topics among Women's National Basketball Association-related Twitter conversation over the course of the 2020 season. Quantitative (cv and umass scores) and qualitative (two expert reviews) approaches were utilized to delineate topic configurations. Marginal topic distance established topic importance. Results from 118,518 tweets revealed 18 conversation topics spanning two overarching themes: social justice issues and on-court performance. The range and depth of the results highlight the importance of the unsupervised topic modeling method (without semi-supervised predetermined topic leads) for considering holistic rather than subsampled or snapshot datasets. This empirical investigation extends the conversation surrounding natural language processing to sport management research and practice, delivers a foundation for unsupervised LDA application to sport consumer conversation, and explores social media conversations during a critical moment for the WNBA.

An exploratory clinical trial of preoperative non-invasive localization before breast-conserving surgery using augmented reality technology

PurposeThis single-center, randomized, prospective, exploratory clinical trial was conducted to assess the clinical efficacy of an augmented reality (AR)—based breast cancer localization imaging solution for patients with breast cancer.MethodsThis clinical trial enrolled 20 women who were diagnosed with invasive breast cancer between the ages of 19 and 80, had a single lesion with a diameter ≥ 5 mm but ≤ 30 mm, had no metastases to other organs, and had not received prior chemotherapy. All patients underwent mammography, ultrasound, computed tomography, and magnetic resonance imaging for preoperative assessment. Patients were randomly assigned to ultrasound-guided skin marking localization (USL) and AR-based localization (ARL) groups (n = 10 in each group). Statistical comparisons between USL and ARL groups were made based on demographics, radiologic features, pathological outcomes, and surgical outcomes using chi-square and Student t-tests.ResultsTwo surgeons performed breast-conserving surgery on 20 patients. Histopathologic evaluation of all patients confirmed negative margins. Two independent pathologists evaluated the marginal distances, and there were no intergroup differences in the readers' estimates (R1, 6.20 ± 4.37 vs. 5.04 ± 3.47, P = 0.519; R2, 5.10 ± 4.31 vs. 4.10 ± 2.38, P = 0.970) or the readers' average values (5.65 ± 4.19 vs. 4.57 ± 2.84, P = 0.509). In comparing the tumor plane area ratio, there was no statistically significant difference between the two groups in terms of either reader's mean values (R1, 15.90 ± 9.52 vs. 19.38 ± 14.05, P = 0.525; R2, 15.32 ± 9.48 vs. 20.83 ± 12.85, P = 0.290) or the overall mean values of two readers combined (15.56 ± 9.11 vs. 20.09 ± 13.38, P = 0.388). Convenience, safety, satisfaction, and reusability were all superior in the AR localization group (P < 0.001) based on the two surgeons' responses.ConclusionAR localization is an acceptable alternative to ultrasound-guided skin marking with no significant differences in surgical outcomes.

Abstract PO1-23-05: Exploratory clinical trial of preoperative non-invasive localization before surgery in breast cancer patients using augmented reality technology

Abstract Purpose: A single-center, randomized, prospective exploratory clinical trial to assess the clinical efficacy of augmented reality(AR)-based breast cancer localization medical imaging solution in patients with breast cancer Methods: This prospective exploratory clinical trial enrolled 20 women who were diagnosed with invasive breast cancer between the ages of 19 and 70, had a single lesion or multifocal lesion with a lesion size of 5mm or more and 30mm or less, had no metastasis to other organs, and had not received prior chemotherapy. All patients underwent mammography, ultrasound, CT, and MRI for preoperative assessment. Patients were randomly assigned to US-guided skin marking (control) and SKIA-breast (AR localization) groups of 10 each. Result: Two surgeons performed breast-conserving surgery on twenty patients. Pathologic evaluation of all patients confirmed negative margins. Two independent pathologists evaluated marginal distance, and neither the two reader's estimates (R1, 6.20±4.37 vs. 5.04±3.47, p=0.519; R2, 5.10±4.31 vs. 4.10±2.38, p= 0.970) nor the two reader's average values (5.65±4.19 vs. 4.57±2.84, p= 0.509) showed any difference between the two groups. In comparing the tumor plane area ratio, there was no statistically significant difference between the two groups for the two readers (R1, 15.90±9.52 vs. 19.38±14.05, p=0.525; R2, 15.32±9.48 vs. 20.83±12.85, p=0.290) and the two reader's mean values (15.56±9.11 vs. 20.09±13.38, p=0.388). Based on the two surgeons' responses, convenience, safety, satisfaction, and reusability were all superior in the AR localization group (p=0.000). Conclusion: AR localization is an acceptable alternative to US-guided skin marking with no significant differences in surgical outcomes. Citation Format: Minah Lee, Jun Woo Lee, Woosung Lim, Joohyun Woo, Hyun-Goo Kim, Se Hyun Paek, SangHui Park, Ji Min Kim, Jin Chung, Jee Eun Lee, Jeoung Hyun Kim. Exploratory clinical trial of preoperative non-invasive localization before surgery in breast cancer patients using augmented reality technology [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-23-05.

An Efficient Knowledge Transfer Strategy for Spiking Neural Networks from Static to Event Domain

Spiking neural networks (SNNs) are rich in spatio-temporal dynamics and are suitable for processing event-based neuromorphic data. However, event-based datasets are usually less annotated than static datasets. This small data scale makes SNNs prone to overfitting and limits their performance. In order to improve the generalization ability of SNNs on event-based datasets, we use static images to assist SNN training on event data. In this paper, we first discuss the domain mismatch problem encountered when directly transferring networks trained on static datasets to event data. We argue that the inconsistency of feature distributions becomes a major factor hindering the effective transfer of knowledge from static images to event data. To address this problem, we propose solutions in terms of two aspects: feature distribution and training strategy. Firstly, we propose a knowledge transfer loss, which consists of domain alignment loss and spatio-temporal regularization. The domain alignment loss learns domain-invariant spatial features by reducing the marginal distribution distance between the static image and the event data. Spatio-temporal regularization provides dynamically learnable coefficients for domain alignment loss by using the output features of the event data at each time step as a regularization term. In addition, we propose a sliding training strategy, which gradually replaces static image inputs probabilistically with event data, resulting in a smoother and more stable training for the network. We validate our method on neuromorphic datasets, including N-Caltech101, CEP-DVS, and N-Omniglot. The experimental results show that our proposed method achieves better performance on all datasets compared to the current state-of-the-art methods. Code is available at https://github.com/Brain-Cog-Lab/Transfer-for-DVS.

Functional Data Clustering Method Based on Shape Information and Functional Mahalanobis Distance

Function-type data can provide insights into the internal structure of the data and facilitate the extraction of data features from the perspective of interactive derivative functions. The paper proposes a nonparametric clustering method for function-based data that incorporates first-order and second-order derivative function information into the Marginal distance for clustering function-based data. The method is based on the traditional K-means algorithm and is designed to cluster function-based data that contains rich shape information. The paper evaluated the performance of the algorithm of the text by comparing its purity and adjusted Rand coefficients against six other clustering algorithms on three different datasets. The results show that the algorithm of the text outperforms the other algorithms, demonstrating its outstanding performance, wide applicability, and practical significance in solving real-world problems.

A new spectral distance based on adaptive selection algorithm for non-stationary time series

We propose a distance metric to quantify the dissimilarity between time series from the perspective of the frequency domain, especially non-stationary time series, called the Wasserstein Hilbert marginal spectral distance (WHMS). It calculates the Wasserstein distance between the Hilbert marginal spectral (HMS) of the time series, where HMS is the integral of Hilbert spectrum. Hilbert spectrum is obtained by the Hilbert transform performed on the intrinsic mode functions (IMFs) generated after the Empirical Mode Decomposition (EMD) operation. On this basis, we propose an IMF adaptive selection algorithm to improve the feature extraction accuracy of the HMS, free from a large number of prior experiments to set the threshold. We demonstrate that WHMS distance can be used as a general measure of time series from both theoretical and practical aspects. Combined with multidimensional scaling (MDS), we comprehensively evaluate the proposed method through simulation experiments and empirical data. The results confirm the good applicability of the WHMS distance combined with the IMF adaptive selection algorithm, which can distinguish different types of complex systems, is robust to noise, and is superior to Wasserstein–Fourier (WF) distance, Wasserstein–Fourier distance with short time Fourier transform (STFT), Euclidean distance and Chebyshev distance.

Prediction of gap balancing based on 2-D radiography in total knee arthroplasty for knee osteoarthritis patients

BackgroundTo investigate the influence of osteophytes on postoperative gap balancing, and to work out a predictive model of the relationship between osteophyte size and gap gaining in primary total knee replacement.MethodsOne hundred and ten patients were enrolled in the study. Pre- and postoperative radiographs were collected and analyzed. They were assigned to the training dataset and test dataset randomly at a ratio of 9:1 by using the statistical package R (version 4.0.5). Size and marginal distances of osteophytes, planned bone cut planes, predicted bone cuts and joint gaps were labeled on the preoperative standing anteroposterior and lateral views, while actual bone cuts and joint gaps were recorded on the postoperative plain films, respectively. Statistical analysis was performed.ResultsActual joint gaps were significantly related to the distances of medial and lateral predictive bone cutting lines, bone cut thickness on tibial side and posterior condylar, as well as size and marginal distances of osteophytes (P < 0.05). A predictive equation was generated, with a root mean square error (RMSE) of 3.4761 in validation. A 2-D planning system with adjustable input parameters and dim predictive outputs on joint gap was developed. The equation is S left(Joint Gapright)=1.82+0.15*y+0.552*Tibial cut+0.953*Femoral cut+0.197*Post CondyleConclusionPostoperative joint gap can be predicted on the basis of preoperative measurements on 2-D plain films. Larger sample size may help improve the effectiveness and accuracy of the predictive equation.

Association between distance to nearest automated external defibrillator and socioeconomic deprivation in Great Britain

Abstract Background Disparities in socioeconomic status manifest in death from out of hospital cardiac arrest (OHCA). Inequalities in access to automated external defibrillators (AED) due to socioeconomically determined location of AEDs may be associated with OHCA survival. Purpose To investigate distances between nearest AED location and postcode by socioeconomic deprivation for the population of Great Britain (GB). Methods AED location data for 2022 were drawn from the British Heart Foundation The Circuit registry. All 1,677,466 unit postcodes in Great Britain were linked to the Index of Deprivation for the corresponding Lower Layer Super Output Area using the 2019 Ministry of Housing Communities and Local Government tool. The median distance by road from the centre of each postcode to the nearest AED was calculated using geographic information systems. The proportion available 24/7 was also reported. Multilevel linear mixed regression models fit by country and adjusted for urban/rural status and clustering to calculate predicted marginal mean distances to the nearest active AED, and nearest AED accessible 24/7 by deprivation decile. Results 78,425 unique AED locations were included in the analysis (Table 1). The median distance to nearest AED was 762.1m (IQR 410.8 - 1221.0) in GB; 742.7m (407.2 - 1340.7) in Scotland, 738.8m (422.6 - 1223.5) in England, and 511.5m (282.9 - 937.3) in Wales. For 24/7 access AEDs, median distances were greater: 964.0m (524.7 - 1667.5), 994.1m (538.5 – 1832.3), 990.8m (544.9 - 1683.3), and 569.7m (316.8 - 1033.4), respectively. Overall, there was little evidence of a trend in nearest AED distance by deprivation (fig 1). In Scotland the nearest AED was, on average, 108m further away in the most compared with the least deprived areas. In England, the nearest AED was closer to postcodes in the least deprived areas of England (predicted mean 20m) and Wales (predicted mean 95m). In England, AEDs with 24/7 access were further from postcodes with increasing deprivation (predicted mean 99m further than least deprived areas). In Scotland, nearest 24/7 access AEDs were 317m further away in the most deprived areas, and 62m closer in Wales, but with no evidence of a linear trend. At a typical walking speed (1.31m/sec), on average in GB it would take 17 minutes to retrieve an AED and 21 minutes out of hours. In England and Scotland out of hours, those in the most deprived areas would be delayed access to an AED by 2.5 and 8 minutes respectively compared the least deprived. Conclusions Overall, the median distance to nearest AED across postcodes in GB is at odds with the likelihood of successful resuscitation. Graded social inequalities were not observed in AED access across GB. However, substantive differences were observed in 24/7 AED access between most and least deprived areas of England and Scotland. Greater and more equitable access to AEDs has the potential to save lives and improve outcomes for people with OHCA.

Influence of Direct and Indirect Digital Data Capturing Techniques on Marginal Accuracy and Internal Fit of CAD/CAM Fabricated PEEK and Zirconia Crowns

Purpose: The aim of this study was to evaluate the marginal and internal fitness of PEEK and Zirconia crowns fabricated by three different impression techniques. Materials and methods: Thirty standardized acrylic die models were prepared. Dies were divided into two main groups (n= 15): Group 1: CAD/CAM fabricated zirconia crowns. Group 2: CAD/CAM fabricated PEEK crowns. Each group was subdivided into three subgroups (n= 5), according to the impression technique using for data acquisition: Sub-group (A): direct digital scans of the model used intraoral digital scanner. Sub-group (B): indirect digital scans of the silicon impressions using extraoral scanner. Sub-group (C): indirect digital scans of the stone casts obtained from silicon impressions using extraoral scanner. Crowns were then designed and fabricated. Marginal and internal gaps were measured using silicone replica technique and field digital microscope. Results: Independent t-test was used for comparison between groups. Comparisons between subgroups were performed by one-way analysis of variance (ANOVA) test, followed by Bonferroni’s post hoc test. In Zirconia group, the highest mean value was recorded in stone cast scan group, followed by intraoral scan, with the least value recorded in impression scan subgroup. In PEEK group, the highest mean value was recorded in stone cast group, followed by impression scan, with the least value recorded in intraoral scan subgroup. Conclusions: All tested crowns showed marginal gap distance values within the clinically acceptable levels. PEEK crowns had lower vertical marginal gap distance than zirconia crowns with different scanning techniques.

Influence of Direct and Indirect Digital Data Capturing Techniques on Marginal Accuracy and Internal Fit of CAD/CAM Fabricated PEEK and Zirconia Crowns

Purpose: The aim of this study was to evaluate the marginal and internal fitness of PEEK and Zirconia crowns fabricated by three different impression techniques. Materials and methods: Thirty standardized acrylic die models were prepared. Dies were divided into two main groups (n= 15): Group 1: CAD/CAM fabricated zirconia crowns. Group 2: CAD/CAM fabricated PEEK crowns. Each group was subdivided into three subgroups (n= 5), according to the impression technique using for data acquisition: Sub-group (A): direct digital scans of the model used intraoral digital scanner. Sub-group (B): indirect digital scans of the silicon impressions using extraoral scanner. Sub-group (C): indirect digital scans of the stone casts obtained from silicon impressions using extraoral scanner. Crowns were then designed and fabricated. Marginal and internal gaps were measured using silicone replica technique and field digital microscope. Results: Independent t-test was used for comparison between groups. Comparisons between subgroups were performed by one-way analysis of variance (ANOVA) test, followed by Bonferroni’s post hoc test. In Zirconia group, the highest mean value was recorded in stone cast scan group, followed by intraoral scan, with the least value recorded in impression scan subgroup. In PEEK group, the highest mean value was recorded in stone cast group, followed by impression scan, with the least value recorded in intraoral scan subgroup. Conclusions: All tested crowns showed marginal gap distance values within the clinically acceptable levels. PEEK crowns had lower vertical marginal gap distance than zirconia crowns with different scanning techniques.

The proximity between styloid process and internal carotid artery as a possible risk factor for dissection: a case–control study

PurposeThe anatomical proximity of the styloid process (SP) to the ipsilateral internal carotid artery (ICA) has been recently recognized as a possible risk factor for carotid artery dissection (CAD). We aimed to verify this hypothesis by comparing the minimum distance between SP and ICA in young adult patients (< 55 years) with and without CAD.MethodsThirty-one CAD patients (cases) were compared with 41 sex-matched patients without dissection, group one of control (G1), and with 16 sex-matched patients with vertebral artery dissection (VAD), group two of control (G2). Two independent observers measured, on CT angiography images, the minimum distance on the axial plane between the SP and ICA in cases and controls. They evaluated both the intercentric and the marginal distance. Differences between groups were estimated by Student t-test.ResultsSP-ICA intercentric distance ipsilateral to dissection was significantly shorter compared to that of the contralateral side of cases (p < 0.001), to those of left and right side of G1 patients (p < 0.001 for both), and to those of left and right side of G2 patients (p < 0.001 for both). SP-ICA marginal distance of cases was significantly shorter compared to those of left and right side of G1 patients (p < 0.001 for both) and to those of left and right side of G2 patients (p < 0.001 for both).ConclusionA short SP-ICA distance appears to be a risk factor for CAD as it likely induces a continuous microtraumatism of the vessel wall during normal head and neck movements.

Marginal gap distance and cyclic fatigue loading for different all-ceramic endocrowns.

The aim of the study was to evaluate the effect of the fabrication techniques of two types of glass ceramics on the marginal gap distance and the fracture resistance of endocrown restorations after cyclic loading. Forty extracted mandibular first molars were root canal treated. Decoronation was done for all the endodontically treated teeth 2 mm above the cemento-enamel junction. The teeth were individually fixed vertically into epoxy resin mounting cylinders. All teeth were prepared to receive endocrown restorations. The prepared teeth were randomly divided into four equal groups (n=10) according to the all-ceramic materials and technique used for endocrown construction as follows: Group I (n=10): Pressable lithium disilicate glass ceramics (IPS e-max Press), Group II (n=10): Pressable zirconia-reinforced lithium disilicate glass ceramics (Celtra Press), Group III (n=10): Machinable lithium disilicate glass ceramics (IPS e-max CAD), Group IV (n=10): Machinable zirconia-reinforced lithium disilicate glass ceramics (Celtra Duo). The endocrowns were cemented using dual-cure resin cement. All endocrowns were subjected to fatigue loading. The cycles were repeated 120,000 times to clinically simulate one year chewing condition. Marginal gap distance of all endocrowns was measured directly using a digital microscope with x100 magnification. The load required to failure was recorded in Newton. Data were collected, tabulated, and statistically analyzed. Fracture resistance testing of all-ceramic crowns revealed a statistically significant difference between all different ceramic materials used in this study (p-value <0.001). On the other hand, there was a statistically significant difference between all the four ceramic crowns for the marginal gap distance either before or after fatigue cyclic loading. After considering the limitation of the current study, the following conclusions were given: endocrowns are considered one of the promising minimally invasive restorations for root canal treated molars. CAD/CAM technology revealed better results than heat press technology regarding the fracture resistance of glass ceramics. Heat Press technology revealed better results than CAD/CAM technology regarding the marginal accuracy of glass ceramics.

Comparison of Digital and Conventional Impressions Based on the 3D Fit of Crowns.

To evaluate intraoral scanners, scannable impression materials, and conventional methods for impression-taking with regard to precision and accuracy. Ten impressions per technique were initially taken from a cobalt-chromium (Co-Cr) master model (chamfer preparation for the maxillary right first molar) for fabricating Co-Cr crown copings. The test specimens, their respective metal master model, and the plaster casts from the conventional impressions were then digitized with the light-optical Atos Triple Scanner. The fit of the dental crown copings was measured in two and three dimensions using computer-aided design software. Global differences between the methods were determined using Kruskal-Wallis test, and pairwise comparisons were performed with Mann-Whitney U test. The significance level was set at .05. The 2D analysis showed that the average absolute marginal distance of the crown copings was 0.026 mm when fabricated with the digital impression method and 0.038 mm when fabricated with the conventional method (P = .028). However, the 3D analysis revealed that the conventional group had a smaller marginal gap (0.028 mm) compared to the digital group (0.06 mm; P = .015). This study showed that, in the field of single-crown prosthetics, digital impressions and the use of scannable impression material could serve as alternatives to conventional methods in the future.

Domain Adaptation Preconceived Hashing for Unconstrained Visual Retrieval.

Learning to hash has been widely applied for image retrieval due to the low storage and high retrieval efficiency. Existing hashing methods assume that the distributions of the retrieval pool (i.e., the data sets being retrieved) and the query data are similar, which, however, cannot truly reflect the real-world condition due to the unconstrained visual cues, such as illumination, pose, background, and so on. Due to the large distribution gap between the retrieval pool and the query set, the performances of traditional hashing methods are seriously degraded. Therefore, we first propose a new efficient but transferable hashing model for unconstrained cross-domain visual retrieval, in which the retrieval pool and the query sample are drawn from different but semantic relevant domains. Specifically, we propose a simple yet effective unsupervised hashing method, domain adaptation preconceived hashing (DAPH), toward learning domain-invariant hashing representation. Three merits of DAPH are observed: 1) to the best of our knowledge, we first propose unconstrained visual retrieval by introducing DA into hashing for learning transferable hashing codes; 2) a domain-invariant feature transformation with marginal discrepancy distance minimization and feature reconstruction constraint is learned, such that the hashing code is not only domain adaptive but content preserved; and 3) a DA preconceived quantization loss is proposed, which further guarantees the discrimination of the learned hashing code for sample retrieval. Extensive experiments on various benchmark data sets verify that our DAPH outperforms many state-of-the-art hashing methods toward unconstrained (unrestricted) instance retrieval in both single- and cross-domain scenarios.

Deep Transfer Network with Multi-Space Dynamic Distribution Adaptation for Bearing Fault Diagnosis.

Domain adaptation-based bearing fault diagnosis methods have recently received high attention. However, the extracted features in these methods fail to adequately represent fault information due to the versatility of the work scenario. Moreover, most existing adaptive methods attempt to align the feature space of domains by calculating the sum of marginal distribution distance and conditional distribution distance, without considering variable cross-domain diagnostic scenarios that provide significant cues for fault diagnosis. To address the above problems, we propose a deep convolutional multi-space dynamic distribution adaptation (DCMSDA) model, which consists of two core components: two feature extraction modules and a dynamic distribution adaptation module. Technically, a multi-space structure is proposed in the feature extraction module to fully extract fault features of the marginal distribution and conditional distribution. In addition, the dynamic distribution adaptation module utilizes different metrics to capture distribution discrepancies, as well as an adaptive coefficient to dynamically measure the alignment proportion in complex cross-domain scenarios. This study compares our method with other advanced methods, in detail. The experimental results show that the proposed method has excellent diagnosis performance and generalization performance. Furthermore, the results further demonstrate the effectiveness of each transfer module proposed in our model.

MicroCT evaluation for CAD/CAM occlusal veneer fit using two materials and three cement space settings.

This study was aimed to evaluate the fit of occlusal veneer restoration for two CAD/CAM materials with different cement space settings, using microCT scans. Sixty resin dies were made and divided into two groups (n=30) according to the materials, (I): Hybrid all-ceramic, and (II): zirconia-reinforced lithium silicate glass-ceramic. Each group was subdivided into three subgroups (n=10) according to the cement space parameters (30, 40, and 50 µm). Occlusal veneers for the six subgroups were milled. A circle with 20 different sections was placed at the center of every scanned specimen to measure four different locations (Occlusal, Axial, Marginal, and Absolute marginal discrepancy). Data were analyzed using two-way ANOVA at a 0.05 level of significance. There was no statistically significant effect of material type on the mean values of internal and marginal gaps for the three cement space parameters (P>0.05). There were no statistically significant differences in the occlusal and axial gap between the cement space parameters, furthermore, there were statistically significant differences in marginal gap distances and absolute marginal discrepancies (P>0.05). Hybrid all-ceramic showed smaller marginal and internal discrepancies than zirconia-reinforced lithium silicate glass-ceramic without statistically significant differences, and, for both materials, 50 µm cement space significantly improved the marginal fit and absolute marginal discrepancy.

Real-Time Learning from an Expert in Deep Recommendation Systems with Application to mHealth for Physical Exercises.

Recommendation systems play an important role in today’s digital world. They have found applications in various areas such as music platforms, e.g., Spotify, and movie streaming services, e.g., Netflix. Less research effort has been devoted to physical exercise recommendation systems. Sedentary lifestyles have become the major driver of several diseases as well as healthcare costs. In this paper, we develop a recommendation system to recommend daily exercise activities to users based on their history, profiles and similar users. The developed recommendation system uses a deep recurrent neural network with user-profile attention and temporal attention mechanisms. Moreover, exercise recommendation systems are significantly different from streaming recommendation systems in that we are not able to collect click feedback from the participants in exercise recommendation systems. Thus, we propose a real-time, expert-in-the-loop active learning procedure. The active learner calculates the uncertainty of the recommendation system at each time step for each user and asks an expert for recommendation when the certainty is low. In this paper, we derive the probability distribution function of marginal distance, and use it to determine when to ask experts for feedback. Our experimental results on a mHealth and MovieLens datasets show improved accuracy after incorporating the real-time active learner with the recommendation system.