Stack Architecture Research Articles

Highly Efficient Fully Integrated Multivoltage-Domain Power Management With Enhanced PSR and Low Cross-Regulation

Multivoltage domains are urgently needed in modern system on chips (SoCs), while existing solutions such as switched dc−dc converter, switched capacitor converter, and low-dropout regulator (LDO) do not generate multivoltage domains conveniently and inexpensively. This article presents a highly efficient fully integrated power management strategy to provide the multiple voltage domains. The proposed architecture stacks a main LDO and several auxiliary push−pull regulators (APPRs) to generate multiple voltages for various loads in one SoC chip. The APPR regulates the load current by either absorbing or providing the additional current, which stabilizes the output voltages, increases the power supply rejection (PSR) and decreases the cross-regulation. A prototype with two output terminals is implemented with a standard 0.18- <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">μ</i> m CMOS technology. The whole system achieves high power efficiency of 96.5% and high PSR of −62 and −142 dB for upper and lower outputs, respectively. The chip area is 980 <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">μ</i> m × 500 <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">μ</i> m and the total quiescent current is 239 <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">μ</i> A. In addition, all off-chip components are eliminated, which is favorable for monolithic realization. The reduced system cost and the reduced electromagnetic interference also simplify the power management significantly, which helps to enable low-power and compact SoCs.

The Comparison and Interpretation of Machine-Learning Models in Post-Stroke Functional Outcome Prediction.

Prediction of post-stroke functional outcomes is crucial for allocating medical resources. In this study, a total of 577 patients were enrolled in the Post-Acute Care-Cerebrovascular Disease (PAC-CVD) program, and 77 predictors were collected at admission. The outcome was whether a patient could achieve a Barthel Index (BI) score of >60 upon discharge. Eight machine-learning (ML) methods were applied, and their results were integrated by stacking method. The area under the curve (AUC) of the eight ML models ranged from 0.83 to 0.887, with random forest, stacking, logistic regression, and support vector machine demonstrating superior performance. The feature importance analysis indicated that the initial Berg Balance Test (BBS-I), initial BI (BI-I), and initial Concise Chinese Aphasia Test (CCAT-I) were the top three predictors of BI scores at discharge. The partial dependence plot (PDP) and individual conditional expectation (ICE) plot indicated that the predictors’ ability to predict outcomes was the most pronounced within a specific value range (e.g., BBS-I < 40 and BI-I < 60). BI at discharge could be predicted by information collected at admission with the aid of various ML models, and the PDP and ICE plots indicated that the predictors could predict outcomes at a certain value range.

Accurate prediction of band gap of materials using stacking machine learning model

The prediction of the band gap of semiconductor materials using machine learning has gradually progressed in recent years. However, the performance of such prediction still needs further optimization. This work applies the stacking approach, which fuses the output of multiple baseline models, to further enhance the performance of band gap regression. Ten baseline models are optimized to predict the band gap of materials. Afterwards, the output of models with relatively better performance is used as the input features of the stacking approach. This research employed a benchmark dataset containing 3896 inorganic compounds, with 136 dimensions, and a newly established complex database (E-AFLOW), containing 21,534 compounds with 206 dimensions, to prove the effectiveness of different models. The trained stacking model based on the E-AFLOW database is then applied to determine the band gaps of different new compounds. The results demonstrate that the stacking model has the highest R2 value, at 0.920, in benchmark dataset and a value of 0.917 in the E-AFLOW dataset, with 5-flod cross validation. For the E-AFLOW dataset, the improvement percentage of RMSE, MAE, MAPE, and R2 of the stacking model to GBDT, XGB, RF, and LGB input baseline models are between 3.06%–17.54%, 8.12%–33.25%, 7.69%–33.33%, and 0.66%-4.44%, respectively. In real applications, the trained stacking model based on the E-AFLOW dataset can predict the band gaps of 78.57% of new materials within ± 8.00% of observed measurements. The minimum deviation between the predicted and observed values is −0.02%, and the maximum is 14.27%. These results convincingly demonstrate the excellent performance of stacking approach in band gap regression.

Ambisense polarity of genome RNA of orthomyxoviruses and coronaviruses.

Influenza viruses and coronaviruses have linear single-stranded RNA genomes with negative and positive sense polarities and genes encoded in viral genomes are expressed in these viruses as positive and negative genes, respectively. Here we consider a novel gene identified in viral genomes in opposite direction, as positive in influenza and negative in coronaviruses, suggesting an ambisense genome strategy for both virus families. Noteworthy, the identified novel genes colocolized in the same RNA regions of viral genomes, where the previously known opposite genes are encoded, a so-called ambisense stacking architecture of genes in virus genome. It seems likely, that ambisense gene stacking in influenza and coronavirus families significantly increases genetic potential and virus diversity to extend virus-host adaptation pathways in nature. These data imply that ambisense viruses may have a multivirion mechanism, like "a dark side of the Moon", allowing production of the heterogeneous population of virions expressed through positive and negative sense genome strategies.

Advancement of Chip Stacking Architectures and Interconnect Technologies for Image Sensors

Abstract Numerous technology breakthroughs have been made in image sensor development in the past two decades. Image sensors have evolved into a technology platform to support many applications. Their successful implementation in mobile devices has accelerated market demand and established a business platform to propel continuous innovation and performance improvement extending to surveillance, medical, and automotive industries. This overview briefs the general camera module and the crucial technology elements of chip stacking architectures and advanced interconnect technologies. This study will also examine the role of pixel electronics in determining the chip stacking architecture and interconnect technology of choice. It is conducted by examining a few examples of CMOS image sensors (CIS) for different functions such as visible light detection, single photon avalanche photodiode (SPAD) for low light detection, rolling shutter, and global shutter, and depth sensing and light detection and ranging (LiDAR). Performance attributes of different architectures of chip stacking are overviewed. Direct bonding followed by Via-last through silicon via (Via-last TSV) and hybrid bonding (HB) technologies are identified as newer and favorable chip-to-chip interconnect technologies for image sensor chip stacking. The state-of-the-art ultrahigh-density interconnect manufacturability is also highlighted.

Ensemble Learning based Age Invariant Fea-ture Recognition Using Soft Computing

Face recognition system is a state-of-the-art computer vision application within the artificial intelligence arena. Face recognition is the automated recognition of humans for their names/unique ID. The age invariant face recognition is a challenge task in the field of face recog-nition. In this work, we have introduced a stacked support vector machine where kernel activation of prototype examples is combined in nonlinear ways. The proposed work integrates soft compu-ting-based support vector machine (SVM) with deep SVM. The proposed model uses the implied relation between the variables described above in order to optimize their overall performance. Specifically, our method uses three different stages of complex convolution neural networks that detect and analyze the location of faces position and landmarks. This work has introduced cross-age celebrity dataset (CACD) for both single as well as cross-database enabling the transition of age. The proposed work has been implemented in the MATLAB simulation tool considering CACD dataset. Experimental results indicate that our techniques significantly outperform other strategies across a range of challenging metrics.

Three-Layer Stacked Generalization Architecture With Simulated Annealing for Optimum Results in Data Mining

The combination of different machine learning models to a single prediction model usually improves the performance of the data analysis. Stacking ensembles are one of such approaches to build a high performance classifier that can be applied to various contexts of data mining. This study proposes an enhanced stacking ensemble by collating a few machine learning algorithms with two layered meta classifications to address the limitations of existing stacking architecture to utilize Simulated Annealing Algorithm to optimize the classifier configuration in order to reach the best prediction accuracy. The proposed method significantly outperformed three general stacking ensembles of two layers that have been executed using the meta classifiers utilized in the proposed architecture. These assessments have been statistically proven at a 95% confidence level. The novel stacking ensemble has also outperformed the existing ensembles named; Adaboost algorithm, Gradient boosting algorithm, XGBoost classifier and bagging classifiers as well.

BLSTM and CNN Stacking Architecture for Speech Emotion Recognition

BLSTM and CNN Stacking Architecture for Speech Emotion Recognition

Identifying elevated risk for future pain crises in sickle-cell disease using photoplethysmogram patterns measured during sleep: A machine learning approach.

Transient increases in peripheral vasoconstriction frequently occur in obstructive sleep apnea and periodic leg movement disorder, both of which are common in sickle cell disease (SCD). These events reduce microvascular blood flow and increase the likelihood of triggering painful vaso-occlusive crises (VOC) that are the hallmark of SCD. We recently reported a significant association between the magnitude of vasoconstriction, inferred from the finger photoplethysmogram (PPG) during sleep, and the frequency of future VOC in 212 children with SCD. In this study, we present an improved predictive model of VOC frequency by employing a two-level stacking machine learning (ML) model that incorporates detailed features extracted from the PPG signals in the same database. The first level contains seven different base ML algorithms predicting each subject's pain category based on the input PPG characteristics and other clinical information, while the second level is a meta model which uses the inputs to the first-level model along with the outputs of the base models to produce the final prediction. Model performance in predicting future VOC was significantly higher than in predicting VOC prior to each sleep study (F1-score of 0.43 vs. 0.35, p-value <0.0001), consistent with our hypothesis of a causal relationship between vasoconstriction and future pain incidence, rather than past pain leading to greater propensity for vasoconstriction. The model also performed much better than our previous conventional statistical model (F1 = 0.33), as well as all other algorithms that used only the base-models for predicting VOC without the second tier meta model. The modest F1 score of the present predictive model was due in part to the relatively small database with substantial imbalance (176:36) between low-pain and high-pain subjects, as well as other factors not captured by the sleep data alone. This report represents the first attempt ever to use non-invasive finger PPG measurements during sleep and a ML-based approach to predict increased propensity for VOC crises in SCD. The promising results suggest the future possibility of embedding an improved version of this model in a low-cost wearable system to assist clinicians in managing long-term therapy for SCD patients.

Reservoir stack machines

Memory-augmented neural networks equip a recurrent neural network with an explicit memory to support tasks that require information storage without interference over long times. A key motivation for such research is to perform classic computation tasks, such as parsing. However, memory-augmented neural networks are notoriously hard to train, requiring many backpropagation epochs and a lot of data. In this paper, we introduce the reservoir stack machine, a model which can provably recognize all deterministic context-free languages and circumvents the training problem by training only the output layer of a recurrent net and employing auxiliary information during training about the desired interaction with a stack. In our experiments, we validate the reservoir stack machine against deep and shallow networks from the literature on three benchmark tasks for Neural Turing machines and six deterministic context-free languages. Our results show that the reservoir stack machine achieves zero error, even on test sequences longer than the training data, requiring only a few seconds of training time and 100 training sequences.

Self-Assembly of Functionalized Lipophilic Guanosines into Cation-Free Stacked Guanine-Quartets.

The hierarchical self-assembly of various lipophilic guanosines exposing either a phenyl or a ferrocenyl group in the C(8) position was investigated. In a solution, all the derivatives were found to self-assemble primarily into isolated guanine (G)-quartets. In spite of the apparent similar bulkiness of the two substituents, most of the derivatives form disordered structures in the solid state, whereas a specific 8-phenyl derivative self-assembles into an unprecedented, cation-free stacked G-quartet architecture.

Three Layer Stacked Generalization Architecture with Simulated Annealing for Optimum Results in Data Mining

The combination of different machine learning models to a single prediction model usually improves the performance of the data analysis. Stacking ensembles are one of such approaches to build a high performance classifier that can be applied to various contexts of data mining. This study proposes an enhanced stacking ensemble by collating a few machine learning algorithms with two layered meta classifications to address the limitations of existing stacking architecture to utilize Simulated Annealing Algorithm to optimize the classifier configuration in order to reach the best prediction accuracy. The proposed method significantly outperformed three general stacking ensembles of two layers that have been executed using the meta classifiers utilized in the proposed architecture. These assessments have been statistically proven at a 95% confidence level. The novel stacking ensemble has also outperformed the existing ensembles named; Adaboost algorithm, Gradient boosting algorithm, XGBoost classifier and bagging classifiers as well.

Study on Analog/RF and Linearity Performance of Staggered Heterojunction Gate Stack Tunnel FET

Staggered heterostructure gate stack TFET is proposed. The analog, RF, and linearity performance of the device were studied in an ATLAS TCAD device simulator. The high K material HfO2 was used in gate stacking. The impact of the gate stack width on the analog, RF, and linearity performance was investigated. Analog/RF performance of the staggered heterostructure gate stack TFET were analyzed in terms of figure of merits (FOMs) like transconductance (gm), transconductance generation factor (TGF), voltage gain, electric field, cut off frequency (fT), maximum frequency of oscillation (fmax), and gain band width (GBW). Gate stacking architecture in heterostructure staggered TFET improves the Ion/Ioff ratio and reduces drain induced barrier lowering with respect to greater gate stack width (toxh), diminishing the short channel effects. Improvement in gm voltage gain, fT, fmax, and GBW) were observed as the gate stack width thickened. A fair comparison of FOMs such as VIP2, VIP3, IIP3, IMD3, and 1 dB–compression point (P1 dB) were carried out by varying the gate stack width to investigate the linearity performance. The simulation results reveal that staggered heterostructure gate stack TFET can be a reasonable competitor for low-power applications and in the design of RF circuits covering a wide range in frequencies of RF spectrum.

A Stacked Neural Network-Based Machine Learning Framework to Detect Activities and Falls Within Multiple Indoor Environments Using Wi-Fi CSI

Device-free methods for activity or fall detection using Wi-Fi channel state information (CSI) have become popular in the literature as they are not intrusive to privacy like competing camera-based solutions. However, such methods require significant setup processes. The objective of this letter is to improve upon the current CSI-based systems by proposing a two-stage modular architecture. A stacked neural network is developed that selects which environment or room a person is located within, before engaging a room-level model for activity recognition. This allows machine learning models to be iteratively deployed to multiple environments without retraining previously deployed room-level models.

Lightweight thermal interface materials based on hierarchically structured graphene paper with superior through-plane thermal conductivity

Graphene-based papers have recently triggered considerable interests in developing the application as thermal interface materials (TIMs) for addressing the interfacial heat transfer issue, but their low through-plane thermal conductivity (κ⊥), resulting from the layer-by-layer stacked architecture, limits the direct use as TIMs. Although various hybrid graphene papers prepared by combining the graphene sheets and the thermally conductive insertions have been proposed to solve this problem, achieving a satisfactory κ⊥ higher than that of commercial TIMs (>5 W m−1 K−1) remains challenging. Here, a strategy aimed at the construction of heat pathways along the through-plane direction inside the graphene paper for achieving a high κ⊥ was demonstrated through the simultaneous filtration of graphene sheets with two different lateral sizes. The as-prepared graphene paper presented a hierarchical structure composed of loosely stacked horizontal layers formed by large graphene sheets, intercalated by a random arrangement of small graphene sheets. Due to the heat pathways formed by small graphene sheets along the through-plane direction, the hierarchically structured graphene paper exhibited an improved κ⊥ as high as 12.6 W m−1 K−1 after a common graphitization post-treatment. In the practical test, our proposed paper as an all-graphene TIM achieved an enhancement in cooling efficiency of ≈ 2.2 times compared to that of the state-of-the-art TIM, demonstrating its superior performance to meet the ever-increasing heat dissipation requirement.

A Numerical Investigation of Stacked Oxide Junctionless High K with Vaccum Metal Oxide Semiconductor Field Effect Transistor

The amalgamation of high dielectric permittivity material along with vaccum based dual material junctionless transistor with surrounding gate (DUM-HIK-VAC-JLSG-MOSFET) was investigated in this paper. The diverse oxide gate stack architecture of DUM-HIK-VAC-JLSG-MOSFET abate various effects arise due to channel shrinking. The evaluation of various parameters such as potential distribution in the channel region, electric field (transverse, lateral), transconductance, electron current density and drain current has been analyzed and a comparative statement between DUM-HIK-VAC-JLSG-MOSFET and the single material gate junctionless high k and vaccum double gate MOSFET(SIM-HIK-VAC-JLSG-MOSFET). The commingle of high dielectric permittivity material with vaccum as gate oxide stack and junctionless transistor with surrounding gate structure Dual material achieve a foremost responsibility to mitigate the device shrinking effects, impel to raise the ION current (10−3A/μm) and diminish the leakage current IOFF of 10−18A/μm to boost the ION/IOFF ratio as 1015. To guarantee the scalability and responsiveness of the device simulations has been done with Synopsys TCAD simulator tool.

Computational Prediction of Cervical Cancer Diagnosis Using Ensemble-Based Classification Algorithm

AbstractCervical cancer is one of the most common cancers among women in the world. As at the earlier stage, cervical cancer has fewer symptoms. Cancer research is vital as the prognosis of cancer enables clinical applications for patients. In this study, we demonstrate a new approach that applies an ensemble approach to machine learning models for the automatic diagnosis of cervical cancer. The dataset used in the study is the cervical cancer dataset available at the University of California Irvine database repository. Initially, missing values are imputed (k-nearest neighbors) and then the data are balanced (oversampled). Two feature selection approaches are used to extract the most significant features. The proposed stacking architecture, applied for the first time on the cervical cancer dataset, used time elapse of 5.6 s and achieved an area under the curve score of 99.7% performing better than the methods used in previous works. The objective of the study is to propose a computational model that can predict the diagnosis of cervical cancer efficiently. Further, the proposed learning architecture is gauged with several ensemble approaches like random forest, gradient boosting, voting ensemble and weighted voting ensemble to perceive the enhancement.

Breath analysis based early gastric cancer classification from deep stacked sparse autoencoder neural network

Deep learning is an emerging tool, which is regularly used for disease diagnosis in the medical field. A new research direction has been developed for the detection of early-stage gastric cancer. The computer-aided diagnosis (CAD) systems reduce the mortality rate due to their effectiveness. In this study, we proposed a new method for feature extraction using a stacked sparse autoencoder to extract the discriminative features from the unlabeled data of breath samples. A Softmax classifier was then integrated to the proposed method of feature extraction, to classify gastric cancer from the breath samples. Precisely, we identified fifty peaks in each spectrum to distinguish the EGC, AGC, and healthy persons. This CAD system reduces the distance between the input and output by learning the features and preserve the structure of the input data set of breath samples. The features were extracted from the unlabeled data of the breath samples. After the completion of unsupervised training, autoencoders with Softmax classifier were cascaded to develop a deep stacked sparse autoencoder neural network. In last, fine-tuning of the developed neural network was carried out with labeled training data to make the model more reliable and repeatable. The proposed deep stacked sparse autoencoder neural network architecture exhibits excellent results, with an overall accuracy of 98.7% for advanced gastric cancer classification and 97.3% for early gastric cancer detection using breath analysis. Moreover, the developed model produces an excellent result for recall, precision, and f score value, making it suitable for clinical application.

Graphene‐Oriented Construction of 2D SnS for Methanol Gas‐Sensor Application

Group‐IV monochalcogenides with unique 2D puckered honeycomb structure (MX, M = Ge, Sn/X = S, Se) have a great prospect in the gas‐sensor field. Herein, an exploratory study on 2D SnS is synthesized through a solvothermal method, is performed for promising methanol sensor. To break through the inhibition of the general agglomeration and poor conductivity of solvothermal synthesized SnS on gas‐sensing performance, a hybrid of SnS@rGO is designed and constructed based on poly(diallyl dimethyl ammonium chloride) (PDDA)‐assisted electrostatic self‐assembly. With directed growth of SnS on the PDDA‐modulated graphene oxide (GO), an expected architecture characterized by well‐dispersed loosely arranged SnS nanosheets attaching on the highly conductive rGO is achieved. The unique hybrid architecture and the conductivity modulation of rGO promotes sensing responses and dynamic characteristic of SnS‐based sensor toward methanol detection. The achieved SnS@rGO hybrid with less stacking architecture of SnS displays a ≈3.3‐fold enhancement in gas response to 15 ppm methanol in comparison to the flower‐like stacked SnS. The formation mechanism of SnS@rGO hybrid based on the oriented growth of SnS by PDDA‐modulated GO is clarified schematically and the response enhancement of the hybrid is demonstrated in terms of the unique architecture and the conductivity modulation.

Various Structural Design Modifications: para-Substituted Diphenylphosphinopyridine Bridged Cu(I) Complexes in Organic Light-Emitting Diodes.

The well-known system of dinuclear Cu(I) complexes bridged by 2-(diphenylphosphino)pyridine (PyrPhos) derivatives Cu2X2L3 and Cu2X2LP2 (L = bridging ligand, P = ancillary ligand) goes along with endless variation options for tunability. In this work, the influence of substituents and modifications on the phosphine moiety of the NP-bridging ligand was investigated. In previous studies, the location of the lowest unoccupied molecular orbital (LUMO) of the copper complexes of the PyrPhos family was found to be located on the NP-bridging ligand and enabled color tuning in the whole visible spectrum. A multitude of dinuclear Cu(I) complexes based on the triple methylated 2-(bis(4-methylphenyl)phosphino)-4-methylpyridine (Cu-1b-H, Cu-1b-MeO, and Cu-1b-F) up to complexes bearing 2-(bis(4-fluorophenyl)phosphino)pyridine (Cu-6a-H) with electron-withdrawing fluorine atoms over many other variations on the NP-bridging ligands were synthesized. Almost all copper complexes were confirmed via single crystal X-ray diffraction analysis. Besides theoretical TDDFT-studies of the electronic properties and photophysical measurements, the majority of the phosphino-modified Cu(I) complexes was tested in solution-processed organic light-emitting diodes (OLEDs) with different heterostructure variations. The best results of the OLED devices were obtained with copper emitter Cu-1b-H in a stack architecture of ITO/PEDOT-PSS (50 nm)/poly-TPD (15 nm)/20 wt % Cu(I) emitter:CBP:TcTA(7:3) (45 nm)/TPBi (30 nm)/LiF(1 nm)/Al (>100 nm) with a high brightness of 5900 Cd/m2 and a good current efficiency of 3.79 Cd/A.