Showing papers by "Samsung published in 2022"

Automatic Detection and Classification System of Domestic Waste via Multimodel Cascaded Convolutional Neural Network

Abstract: Domestic waste classification was incorporated into legal provisions recently in China. However, relying on manpower to detect and classify domestic waste is highly inefficient. To that end, in this article, we propose a multimodel cascaded convolutional neural network (MCCNN) for domestic waste image detection and classification. MCCNN combined three subnetworks (DSSD, YOLOv4, and Faster-RCNN) to obtain the detections. Moreover, to suppress the false-positive predicts, we utilized a classification model cascaded with the detection part to judge whether the detection results are correct. To train and evaluate MCCNN, we designed a large-scale waste image dataset (LSWID), containing 30 000 domestic waste multilabeled images with 52 categories. To the best of our knowledge, the LSWID is the largest dataset on domestic waste images. Furthermore, a smart trash can is designed and applied to a Shanghai community, which helped to make waste recycling more efficient. Experimental results showed a state-of-the-art performance, with an average improvement of 10% in detection precision.

A Fast Droop-Recovery Event-Driven Digital LDO With Adaptive Linear/Binary Two-Step Search for Voltage Regulation in Advanced Memory

Abstract: This letter presents an event-driven digital low-dropout regulator (DLDO) with an adaptive linear/binary two-step search achieving a fast transient response. A two-dimensional (2-D) circular shifting register (CSR) offers an adaptive linear-search regulation. When a large voltage droop occurs, the CSR activates a fast-tracking mode that provides immediate recovery from the droop. Once the linear search by the CSR is completed, a subrange successive-approximation register (Sub-SAR) conducts the binary-search regulation. The full-scale current range of the Sub-SAR is adaptively scaled by referencing the CSR, which reduces the number of searching steps and improves undershoot or overshoot caused by the binary-search operation. Ring amplifier based 1.5b continuous-time (CT) comparators and the asynchronous controllers realize the event-driven operation that breaks a tradeoff between transient response and sampling clock frequency. The proposed DLDO was fabricated in a 40 nm CMOS process. The DLDO can operate in an input voltage V IN range from 0.6 to 1.2 V. When a load current step of 104.2 mA/1 ns was applied at a V IN of 1.0 V, a droop-recovery time and a settling time were measured as 6 and 15 ns, respectively.

EdgeViTs: Competing Light-Weight CNNs on Mobile Devices with Vision Transformers

Abstract: Self-attention based models such as vision transformers (ViTs) have emerged as a very competitive architecture alternative to convolutional neural networks (CNNs) in computer vision. Despite increasingly stronger variants with ever higher recognition accuracies, due to the quadratic complexity of self-attention, existing ViTs are typically demanding in computation and model size. Although several successful design choices (e.g., the convolutions and hierarchical multi-stage structure) of prior CNNs have been reintroduced into recent ViTs, they are still not sufficient to meet the limited resource requirements of mobile devices. This motivates a very recent attempt to develop light ViTs based on the state-of-the-art MobileNet-v2, but still leaves a performance gap behind. In this work, pushing further along this under-studied direction we introduce EdgeViTs, a new family of light-weight ViTs that, for the first time, enable attention based vision models to compete with the best light-weight CNNs in the tradeoff between accuracy and on-device efficiency. This is realized by introducing a highly cost-effective local-global-local (LGL) information exchange bottleneck based on optimal integration of self-attention and convolutions. For device-dedicated evaluation, rather than relying on inaccurate proxies like the number of FLOPs or parameters, we adopt a practical approach of focusing directly on on-device latency and, for the first time, energy efficiency. Extensive experiments on image classification, object detection and semantic segmentation validate high efficiency of our EdgeViTs when compared to the state-of-the-art efficient CNNs and ViTs in terms of accuracy-efficiency tradeoff on mobile hardware. Specifically, we show that our models are Pareto-optimal when both accuracy-latency and accuracy-energy tradeoffs are considered, achieving strict dominance over other ViTs in almost all cases and competing with the most efficient CNNs. Code is available at https://github.com/saic-fi/edgevit .

Drug-likeness scoring based on unsupervised learning

Abstract: Drug-likeness prediction is important for the virtual screening of drug candidates. It is challenging because the drug-likeness is presumably associated with the whole set of necessary properties to pass through clinical trials, and thus no definite data for regression is available. Recently, binary classification models based on graph neural networks have been proposed but with strong dependency of their performances on the choice of the negative set for training. Here we propose a novel unsupervised learning model that requires only known drugs for training. We adopted a language model based on a recurrent neural network for unsupervised learning. It showed relatively consistent performance across different datasets, unlike such classification models. In addition, the unsupervised learning model provides drug-likeness scores that well separate distributions with increasing mean values in the order of datasets composed of molecules at a later step in a drug development process, whereas the classification model predicted a polarized distribution with two extreme values for all datasets presumably due to the overconfident prediction for unseen data. Thus, this new concept offers a pragmatic tool for drug-likeness scoring and further can be applied to other biochemical applications.

The Psychological Well-Being of Older Chinese Immigrants in Canada amidst COVID-19: The Role of Loneliness, Social Support, and Acculturation

Abstract: This study examined the effects of loneliness, social support, and acculturation on psychological well-being, as indexed by general emotional well-being and life satisfaction, of older Chinese adults living in Canada during the COVID-19 pandemic. A total of 168 older Chinese adults, recruited via WeChat and the internet, completed an online study through a facilitated Zoom or phone meeting, or through a website link, individually or in a group. The testing package included demographic information, The UCLA Loneliness Scale, The Multidimensional Perceived Social Support Scale, Vancouver Index of Acculturation, The Satisfaction with Life Scale, and The World Health Organization's Five Well-Being Index. The results showed that the psychological well-being (both general emotional well-being and cognitively perceived life satisfaction) was positively predicted by perceived social support but negatively predicted by loneliness. Acculturation was not predictive of both outcomes, and it did not moderate the predictive relationships of social support or loneliness. The results shed light on the importance of community services that target enhancing social support and reducing loneliness in promoting psychological well-being of older Chinese immigrants in Canada amidst and post the pandemic.

Convolutional Encoder–Decoder Architecture for Speech Enhancement

Abstract: Signal processing faces the quandary of not being able to separate non-stationary noise from speech signal. Traditional methodologies relied on spectral subtraction for the same; however, such techniques relied on approximation of spectral mask of the noise. This paper proposes an effective and novel convolutional encoder–decoder architecture to effectuate clean speech from the input audio through denoising the audio input. The architecture uses skip connections to increase information flow from encoder to decoder, which helped the authors bolster the performance of the network. The generated output is evaluated on objective and subjective metrics like signal-to-noise ratio (SDR), perceptual evaluation of speech quality (PESQ) and short time objective intelligibility (STOI). The proposed system outperforms the state-of-the-art systems with respect to SDR, PESQ and STOI. The architecture finds applications in various fields such as speech recognition, machine translation and telecommunication.

Identification of an EMT-related Gene Signature Predicting Recurrence in Stage II/III Colorectal Cancer

Abstract: To identify a prognostic significant gene signature for predicting colorectal cancer (CRC) recurrence.Traditional prognostic risk assessment in stage II/III CRC patients remains controversial. Epithelial-mesenchymal transition is thought to be closely related to the malignant progression of tumors. Thus, it is promising to establish a prognostic model based on epithelial-mesenchymal transition-related gene (ERG) signature.We retrospectively analyzed transcriptome profiles and clinical information of 1780 stage II/III CRC patients from 15 public datasets. Coefficient variant analysis was used to select reference genes for normalizing gene expression levels. Univariate, LASSO, and multivariate Cox regression analyses were combined to develop the ERG signature predicting disease-free survival (DFS). The patients were divided into high-risk and low-risk based on the ERG signature recurrence risk score. The survival analysis was performed in different CRC cohorts.The proposed ERG signature contained 7 cancer-related ERGs and 3 reference genes. The ERG signature recurrence risk score was prognostically relevant in all cohorts ( P <0.05) and proved as an independent prognostic factor in the training cohort. In the pooled cohort, high-risk CRC patients exhibited worse DFS ( P <0.0001) and overall survival ( P =0.0058) than low-risk patients. The predictive performance of the ERG signature was superior to Oncotype DX colon cancer. An integrated decision tree and nomogram were developed to improve prognosis evaluation.The identified ERG signature is a promising and powerful biomarker predicting recurrence in CRC patients. Moreover, the presented ERG signature might help to stratify patients according to their tumor biology and contribute to personalized treatment.

Incident duration prediction using a bi-level machine learning framework with outlier removal and intra–extra joint optimisation

Abstract: Predicting the duration of traffic incidents is a challenging task due to the stochastic nature of events. The ability to accurately predict how long accidents will last can provide significant benefits to both end-users in their route choice and traffic operation managers in handling of non-recurrent traffic congestion. This paper presents a novel bi-level machine learning framework enhanced with outlier removal and intra–extra joint optimisation for predicting the incident duration on three heterogeneous data sets collected for both arterial roads and motorways from Sydney, Australia and San-Francisco, U.S.A. Firstly, we use incident data logs to develop a binary classification prediction approach, which allows us to classify traffic incidents as short-term or long-term. We find the optimal threshold between short-term versus long-term traffic incident duration, targeting both class balance and prediction performance while also comparing the binary versus multi-class classification approaches using quantiled duration groups and varying threshold split. Secondly, for more granularity of the incident duration prediction to the minute level, we propose a new intra–extra Joint Optimisation algorithm (IEO-ML) which extends multiple baseline ML models tested against several regression scenarios across the data sets. Final results indicate that: (a) 40–45 min is the best split threshold for identifying short versus long-term incidents and that these incidents should be modelled separately, (b) our proposed IEO-ML approach significantly outperforms baseline ML models in 66% of all cases showcasing its great potential for accurate incident duration prediction. Lastly, we evaluate the feature importance and show that time, location, incident type, incident reporting source and weather at among the top 10 critical factors which influence how long incidents will last. • We propose a novel bi-level framework for predicting the incident durations. • We predict incident duration on three data sets with different road networks. • Short-term and long-term traffic accidents should be modelled separately. • Different incident duration extrapolation scenarios analysed. • Our proposed IEO-ML approach outperformed baseline ML models in 66% of cases.

Synthesis of vertically aligned wafer-scale tantalum disulfide using high-Ar/H2S ratio plasma.

Abstract: Nanostructural modification of two-dimensional (2D) materials has attracted significant attention for enhancing hydrogen evolution reaction (HER) activity. In this study, the nanostructure of TaS2films was controlled by controlling the Ar/H2S gas ratio used in plasma-enhanced chemical vapor deposition (PECVD). At a high Ar/H2S gas ratio, vertically aligned TaS2(V-TaS2) films were formed over a large-area (4 in) at a temperature of 250 °C, which, to the best of our knowledge, is the lowest temperature reported for PECVD. Furthermore, the plasma species formed in the injected gas at various Ar/H2S gas ratios were analyzed using optical emission spectroscopy to determine the synthesis mechanism. In addition, the 4 in wafer-scale V-TaS2was analyzed by x-ray photoelectron spectroscopy, transmission electron microscopy, and atomic force microscopy, and the HER performance of the as-synthesized TaS2fabricated with various Ar/H2S ratios was measured. The results revealed that, depending on the film structure of TaS2, the HER performance can be enhanced owing to its structural advantage. Furthermore, the excellent stability and robustness of V-TaS2was confirmed by conducting 1000 HER cycles and post-HER material characterization. This study provides important insights into the plasma-assisted nanostructural modification of 2D materials for application as enhanced electrocatalysts.

Personalization of Power and Mass Uprisings in Dictatorships

Abstract: Abstract Most major nonviolent civil resistance campaigns target autocratic regimes. Yet, most dictators are toppled by their close supporters, not civilian protesters. Building on theories of strategic interactions between leaders, security agents, and protesters, we make three core claims: first, protesters are relatively less likely to mount a major nonviolent uprising against dictatorships with personalized security forces; secondly, personalized security forces are more likely to repress realized protest; and, thirdly, security force personalization shapes the prospects for success of mass uprisings in promoting democratic transitions. We leverage new data on security force personalization—a proxy for loyal security agents—and major nonviolent protest campaigns to test these expectations. Our theory explains why many dictatorships rarely face mass protest mobilization and why uprisings that are met with violent force often fail in bringing about new democracies.

Validation of Age-adjusted Ideal Sagittal Alignment in Terms of Proximal Junctional Failure and Clinical Outcomes in Adult Spinal Deformity

Abstract: Retrospective study.To validate the age-adjusted ideal sagittal alignment in terms of proximal junctional failure (PJF) and clinical outcomes.It is reported that optimal sagittal correction with regard to the age-adjusted ideal sagittal alignment reduces the risk of PJF development. However, few studies have validated this concept. The age-considered optimal correction is likely to be undercorrection in terms of conventional surgical target, such as pelvic incidence (PI)-lumbar lordosis (LL) within ±9°. Therefore, the clinical impact of age-adjusted sagittal alignment should be evaluated along with radiographic effect.Adult spinal deformity patients, aged 50 years and above, who underwent greater than or equal to four-level fusion to sacrum with a minimum of four years of follow-up data were included in this study. Radiographic risk factors for PJF (including age-adjusted ideal PI-LL) were evaluated with multivariate analyses. Three groups were created based on PI-LL offset between age-adjusted ideal PI-LL and actual actual PI-LL: undercorrection, ideal correction, and overcorrection. Clinical outcomes were compared among the three groups.This study included 194 adult spinal deformity patients. The mean age was 68.5 years and there were 172 females (88.7%). PJF developed in 99 patients (51.0%) at a mean postoperative period of 14.9 months. Multivariate analysis for PJF revealed that only PI-LL offset group had statistical significance. The proportion of patients with PJF was greatest in the overcorrection group followed by the ideal correction and undercorrection groups. Overcorrected patients regard to the age-adjusted ideal alignment showed poorer clinical outcomes than the other patient groups.Overcorrection relative to age-adjusted sagittal alignment increases the risk of PJF development and is associated with poor clinical outcomes. Older patients and those with small PI are likely to be overcorrected in terms of the age-adjusted PI-LL target. Therefore, the age-adjusted alignment should be considered more strictly in these patients.

Amorphous Carbon Films for Electronic Applications

Abstract: While various crystalline carbon allotropes, including graphene, have been actively investigated, amorphous carbon (a-C) thin films have received relatively little attention. The a-C is a disordered form of carbon bonding with a broad range of the CC bond length and bond angle. Although accurate structural analysis and theoretical approaches are still insufficient, reproducible structure-property relationships have been accumulated. As the a-C thin film is now adapted as a hardmask in the semiconductor industry and new properties are reported continuously, expectations are growing that it can be practically used as active materials beyond as a simple sacrificial layer. In this perspective review article, after a brief introduction to the synthesis and properties of the a-C thin films, their potential practical applications are proposed, including hardmasks, extreme ultraviolet (EUV) pellicles, diffusion barriers, deformable electrodes and interconnects, sensors, active layers, electrodes for energy, micro-supercapacitors, batteries, nanogenerators, electromagnetic interference (EMI) shielding, and nanomembranes. The article ends with a discussion on the technological challenges in a-C thin films.