Search

This paper is dedicated to the "50 Years of the Relevance Problem" - a long-neglected topic that begs attention from practical statisticians who are concerned with the problem of drawing inference from large-scale heterogeneous data., Comment: Revised (much-improved) version. The procedure (including all the datasets) is implemented in the R-package LPRelevance

Summary High-dimensional $k$-sample comparison is a common task in applications. We construct a class of easy-to-implement distribution-free tests based on new nonparametric tools and unexplored connections with spectral graph theory. The test is shown to have various desirable properties and a characteristic exploratory flavour that has practical consequences for statistical modelling. Numerical examples show that the proposed method works surprisingly well across a broad range of realistic situations.

We have entered a new era of machine learning (ML), where the most accurate algorithm with superior predictive power may not even be deployable, unless it is admissible under the regulatory constraints. This has led to great interest in developing fair, transparent and trustworthy ML methods. The purpose of this article is to introduce a new information-theoretic learning framework (admissible machine learning) and algorithmic risk-management tools (InfoGram, L-features, ALFA-testing) that can guide an analyst to redesign off-the-shelf ML methods to be regulatory compliant, while maintaining good prediction accuracy. We have illustrated our approach using several real-data examples from financial sectors, biomedical research, marketing campaigns, and the criminal justice system., Comment: Keywords: Admissible machine learning; InfoGram; L-Features; Information-theory; ALFA-testing, Algorithmic risk management; Fairness; Interpretability; COREml; FINEml

We report positive bias temperature instability data in replacement metal gate Ge n-channel metal–oxide–semiconductor field-effect transistors with an in-situ gate stack employing an ultrathin (5 A), stable Al2O3 interlayer (IL) using ultrafast (~microseconds) characterization techniques that ensure recovery artifact-free measurements. Comparison with state-of-the-art GeO x IL is also reported besides establishing correlations between band-edge interface trap density ( ${D}_{\text {it}}$ ), mobility ( $\mu$ ), and threshold voltage ( ${V}_{T}$ ) instability. Ultrafast measure-stress-measure (UF-MSM), ultrafast measure-stress-detrap-measure (UF-MSDM), stress-induced-leakage-current (SILC), direct-current current–voltage (DCIV), split capacitance–voltage ( ${C}$ – ${V}$ ), and low-temperature full conductance techniques along with a compact model demonstrate that: 1) trap generation occurs at IL/high- ${k}$ interface during stress; 2) an increase in $\mu$ with reduction in ${D}_{\text {it}}$ does not guarantee better reliability, i.e., ${V}_{T}$ shift and $\mu$ are uncorrelated due to their dependence on separate regions of the gate stack; 3) contributions to total ${V}_{\text {T}}$ degradation from trapping and generated traps are mutually exclusive; 4) UF-MSDM is a powerful tool to estimate trap generation; and 5) ${V}_{T}$ degradation is directly proportional to high- ${k}$ thickness, varies inversely with IL thickness, and reduces with annealing.

A comprehensive modeling framework is presented to predict the time kinetics of negative bias temperature instability stress and recovery during and after dc and ac stresses and also during mixed dc–ac stress. The model uses uncorrelated contributions from the generation of interface and bulk traps and hole trapping in preexisting bulk traps. Ultrafast measured data at different stresses and recovery biases, temperature, duty cycle and frequency, as well as arbitrary time segments with dynamically varying voltage, frequency, and activity are predicted. The role of nitrogen in the gate insulator is explained. End-of-life degradation is determined under dc and ac use conditions.

A comprehensive nonparametric statistical learning framework, called LPiTrack, is introduced for large-scale eye-movement pattern discovery. The foundation of our data-compression scheme is based on a new Karhunen–Loeve-type representation of the stochastic process in Hilbert space by specially designed orthonormal polynomial expansions. We apply this novel nonlinear transformation-based statistical data-processing algorithm to extract temporal-spatial-static characteristics from eye-movement trajectory data in an automated, robust way for biometric authentication. This is a significant step towards designing a next-generation gaze-based biometric identification system. We elucidate the essential components of our algorithm through data from the second Eye Movements Verification and Identification Competition, organized as a part of the 2014 International Joint Conference on Biometrics.

A physics-based modeling framework is proposed to calculate the threshold voltage shift ( $\Delta \text{V}_{\mathrm {T}})$ in planar high-k metal gate (HKMG) n-MOSFETs for positive bias temperature instability (PBTI). Overall $\Delta \text{V}_{\mathrm { {T}}}$ is estimated using the uncorrelated contributions from the trap generation (TG) and the electron trapping subcomponents. The time evolution of $\Delta \text{V}_{\mathrm { {T}}}$ , measured using an ultrafast measure-stress-measure method during dc and ac stress and after dc stress, is predicted for different experimental conditions. The modeled TG component is verified by independent direct-current I–V method. The proposed model explains PBTI in differently processed HKMG gate stacks.

Degradation in planar high-k metal gate p- and n-channel MOSFETs, respectively, under negative bias temperature instability (NBTI) and positive bias temperature instability (PBTI) stress is studied using different characterization methods. Ultrafast measure stress measure (UF-MSM) method with a measurement delay of a few microseconds is used to characterize the threshold voltage shift ( $\Delta V_{T}$ ). Gated-diode or direct current IV is used to directly estimate the trap generation (TG) during BTI, after correcting for the measurement inconsistencies. BTI experiments are performed under DC stress at different stress bias ( $V_{\mathrm {G-STR}}$ ) and temperature ( ${T}$ ) values also under AC stress at different pulse duty cycle (PDC) and frequency ( ${f}$ ) values. Measured $\Delta V_{T}$ as well as TG show remarkable similarities between NBTI and PBTI stress, under both DC and AC stress. It is shown that TG dominates NBTI and PBTI degradation under both DC and AC stress.

To handle the ubiquitous problem of "dependence learning," copulas are quickly becoming a pervasive tool across a wide range of data-driven disciplines encompassing neuroscience, finance, econometrics, genomics, social science, machine learning, healthcare and many more. Copula (or connection) functions were invented in 1959 by Abe Sklar in response to a query of Maurice Frechet. After 60 years, where do we stand now? This article provides a history of the key developments and offers a unified perspective., Comment: A perspective on "60 years of Copula"

The two key issues of modern Bayesian statistics are: (i) establishing principled approach for distilling statistical prior that is consistent with the given data from an initial believable scientific prior; and (ii) development of a consolidated Bayes-frequentist data analysis workflow that is more effective than either of the two separately. In this paper, we propose the idea of “Bayes via goodness-of-fit” as a framework for exploring these fundamental questions, in a way that is general enough to embrace almost all of the familiar probability models. Several examples, spanning application areas such as clinical trials, metrology, insurance, medicine, and ecology show the unique benefit of this new point of view as a practical data science tool.

There is an overwhelmingly large literature and algorithms already available on "large-scale inference problems" based on different modeling techniques and cultures. Our primary goal in this article is not to add one more new methodology to the existing toolbox but instead (i) to clarify the mystery how these different simultaneous inference methods are connected, (ii) to provide an alternative more intuitive derivation of the formulas that leads to simpler expressions in order (iii) to develop a unified algorithm for practitioners. A detailed discussion on representation, estimation, inference, and model selection is given. Applications to a variety of real and simulated datasets show promise. We end with several future research directions.

Consider a big data multiple testing task, where, due to storage and computational bottlenecks, one is given a very large collection of p-values by splitting into manageable chunks and distributing over thousands of computer nodes. This paper is concerned with the following question: How can we find the full data multiple testing solution by operating completely independently on individual machines in parallel, without any data exchange between nodes? This version of the problem tends naturally to arise in a wide range of data-intensive science and industry applications whose methodological solution has not appeared in the literature to date; therefore, we feel it is necessary to undertake such analysis. Based on the nonparametric functional statistical viewpoint of large-scale inference, started in Mukhopadhyay (2016), this paper furnishes a new computing model that brings unexpected simplicity to the design of the algorithm which might otherwise seem daunting using classical approach and notations., Comment: Revised version

A comprehensive modeling framework involving mutually uncorrelated contribution from interface trap generation and hole trapping in pre-existing, process related gate insulator traps is used to study NBTI degradation in SiON and HKMG p-MOSFETs. The model can predict time evolution of degradation during DC and AC stress, time evolution of recovery after stress, impact of stress and recovery bias and temperature, and impact of several AC stress parameters such as pulse frequency, duty cycle, duration of last pulse cycle (half or full) and pulse low bias. The model can successfully explain experimental data measured using fast and ultra-fast methods in SiON and HKMG devices having different gate insulator processes. The trap generation and trapping sub components of the composite model have been verified by independent experiments. Data published by different groups are reconciled and explained. The model can successfully predict long time DC and AC stress data and has been used to determine device degradation at end of life as EOT is scaled for different HKMG devices. (C) 2014 Elsevier Ltd. All rights reserved.

Bump-hunting or mode identification is a fundamental problem that arises in almost every scientific field of data-driven discovery. Surprisingly, very few data modeling tools are available for automatic (not requiring manual case-by-base investigation), objective (not subjective), and nonparametric (not based on restrictive parametric model assumptions) mode discovery, which can scale to large data sets. This article introduces LPMode--an algorithm based on a new theory for detecting multimodality of a probability density. We apply LPMode to answer important research questions arising in various fields from environmental science, ecology, econometrics, analytical chemistry to astronomy and cancer genomics., Comment: I would like to express my sincere thanks to the Editor and the anonymous reviewers for their in-depth comments, which have greatly improved the manuscript

In this paper, we consider the penalized empirical likelihood (PEL) method of Bartolucci (2007) for inference on the population mean which is a modification of the standard empirical likelihood and employs a penalty based on the Mahalanobis-distance. We derive the asymptotic distributions of the PEL ratio statistic when the dimension of the observations increases with the sample size. Finite sample properties of the method are investigated through a small simulation study.

In kernel discriminant analysis, it is common practice to select the smoothing parameter (bandwidth) based on the training data and use it for classifying all unlabeled observations. But this method of selecting a single scale of smoothing ignores the major issue of model uncertainty. Moreover, in addition to depending on the training sample, a good choice of bandwidth may also depend on the observation to be classified, and a fixed level of smoothing may not work well in all parts of the measurement space. So, instead of using a single smoothing parameter, it may be more useful in practice to study classification results for multiple scales of smoothing and judiciously aggregate them to arrive at the final decision. This paper adopts a Bayesian approach to carry out one such multiscale analysis using a probabilistic framework. This framework also helps us to extend our multiscale method for semi-supervised classification, where, in addition to the training sample, one uses unlabeled test set observations to form the decision rule. Some well-known benchmark data sets are analyzed to show the utility of these proposed methods.

The laminar motion of fluid in microchannels is the necessary criteria for integrated microfluidic system as lab-on-a-chip. Experiments were conducted to investigate laminar flow characteristics of dyed water through photoresist microchannels with square pillars. We found that the pillar dimensions on the channel surfaces have significant impact on the flow rate. The evaluated Reynolds number was less than unity in each microfluidic flow. The compatibility between the pillar sizes and corresponding air-water meniscus movement in microfluidic flow has been reported. [DOI: 10.1380/ejssnt.2009.330]

Dramatic increases in the size and complexity of modern datasets have made traditional "centralized" statistical inference prohibitive. In addition to computational challenges associated with big data learning, the presence of numerous data types (e.g. discrete, continuous, categorical, etc.) makes automation and scalability difficult. A question of immediate concern is how to design a data-intensive statistical inference architecture without changing the basic statistical modeling principles developed for "small" data over the last century. To address this problem, we present MetaLP, a flexible, distributed statistical modeling framework., The purpose of this paper is to answer the question: What is the relevance of small-data-ideas in this big-data world? The bigger question is: Should we make difficult things easy or easy things look difficult? The first option will probably make some impact in the long-run, but the second one will surely earn prestigious journal publications in short-run, IEEE Transactions on Big Data (forthcoming). The first report came out in 2015

The impact of gate insulator processes to achieve deeply scaled interlayer (IL)/high-k (HK) bilayer stacks for sub-20-nm CMOS on negative-bias temperature instability and positive-bias temperature instability is studied. IL scaling is done by novel low-thermal-budget rapid-thermal-process-based ultrathin IL and monolayer IL. Innovative IL top surface treatment enables integration of IL and atomic-layer-deposition-based hafnium oxide HK without vacuum break. Fully integrated stacks show scaling of equivalent oxide thickness down to similar to 6 angstrom, with excellent gate leakage, mobility, and world-class BTI. The mechanism responsible for improved BTI is discussed.

A gated-diode or direct current IV method is used to characterize trap generation (TG) under negative-bias temperature instability (NBTI) and positive-bias temperature instability (PBTI) stress in different planar high- $k$ metal gate p-channel and n-channel MOSFETs, respectively. After correction of the measurement delay, very similar time ( $t_{\mathrm {STR}})$ , voltage ( $V_{{G,\mathrm {STR}}})$ , temperature ( $T$ ), AC pulse duty cycle, and frequency ( $f$ ) dependence of TG is seen for NBTI and PBTI. Measured TG shows power-law time dependence with a time exponent of $n\sim $ 0.16 for NBTI and PBTI and for DC and AC stress. Uncoupled nature of voltage acceleration ( $\Gamma )$ and $T$ activation ( $E_{A})$ is seen. Interlayer scaling has a similar impact on $E_{A}$ and $\Gamma $ for NBTI and PBTI. However, the physical location of TG is shown to be different for NBTI and PBTI stress.

Process impact of negative bias temperature instability (NBTI) is studied in silicon oxynitride (SiON) and high- k metal gate (HKMG) p-MOSFETs. An analytical compact model is used to predict long time degradation. NBTI is shown to be governed by the generation of interface and bulk oxide traps and hole trapping in preexisting traps that are mutually uncorrelated. Experimental evidences are provided to independently verify underlying components. Model parameters are extracted; only a few process-dependent parameters are needed to predict the experimental data from wide range of SiON and HKMG p-MOSFETs at various stress bias and temperature. Similarity between SiON and HKMG devices is highlighted.

The required rate of plasma current rise for setting up ultra-low q/sub /spl alpha// (ULQ) discharges have been quantified for the first time. It Is observed that the rate depends inversely on the value of the mode rational barrier to be crossed. The empirical relation obtained has important consequences as regards setting up of such discharges and design of ULQ devices.

Different physics-based negative bias temperature instability (NBTI) models as proposed in the literature are reviewed, and the predictive capability of these models is benchmarked against experimental data. Models that focus exclusively on hole trapping in gate-insulator-process-related preexisting traps are found to be inconsistent with direct experimental evidence of interface trap generation. Models that focus exclusively on interface trap generation are incapable of predicting ultrafast measurement data. Models that assume strong correlation between interface trap generation and hole trapping in switching hole traps cannot simultaneously predict long-time dc stress, recovery, and ac stress and cannot estimate gate insulator process impact. Uncorrelated contributions from generation and recovery of interface traps, together with hole trapping and detrapping in preexisting and newly generated bulk insulator traps, are invoked to comprehensively predict dc stress and recovery, ac duty cycle and frequency, and gate insulator process impact of NBTI. The reaction-diffusion model can accurately predict generation and recovery of interface traps for different devices and experimental conditions. Hole trapping/ detrapping is modeled using a two-level energy well model.

The impact of the gate insulator process on interlayer (IL) hole traps in IL/high-K dual-layer p-MOSFET gate-stack is studied by physical and electrical measurements along with atomistic simulations. Processes that lead to higher concentrations of Hf and N in IL, measured by angle-resolved X-ray photoelectron spectroscopy, result in higher IL hole traps measured by flicker noise in prestress and verified by atomistic simulations. The influence of these process induced preexisting IL hole traps on parametric degradation of p-MOSFETs during Negative bias temperature instability (NBTI) stress is studied. The mechanism responsible for superior NBTI of thermal IL stack, having lower Hf and N content in the IL as compared with Chem-Ox IL stack, is explained.

This paper formulates a penalized empirical likelihood (PEL) method for inference on the population mean when the dimension of the observations may grow faster than the sample size. Asymptotic distributions of the PEL ratio statistic is derived under different component-wise dependence structures of the observations, namely, (i) non-Ergodic, (ii) long-range dependence and (iii) short-range dependence. It follows that the limit distribution of the proposed PEL ratio statistic can vary widely depending on the correlation structure, and it is typically different from the usual chi-squared limit of the empirical likelihood ratio statistic in the fixed and finite dimensional case. A unified subsampling based calibration is proposed, and its validity is established in all three cases, (i)-(iii). Finite sample properties of the method are investigated through a simulation study., Published in at http://dx.doi.org/10.1214/12-AOS1040 the Annals of Statistics (http://www.imstat.org/aos/) by the Institute of Mathematical Statistics (http://www.imstat.org)

Polymethylmethacrylate (PMMA) microfluidic devices have been fabricated using a hot embossing technique to incorporate micro-pillar features on the bottom wall of the device which when combined with either a plasma treatment or the coating of a diamond-like carbon (DLC) film presents a range of surface modification profiles. Experimental results presented in detail the surface modifications in the form of distinct changes in the static water contact angle across a range from 44.3 to 81.2 when compared to pristine PMMA surfaces. Additionally, capillary flow of water (dyed to aid visualization) through the microfluidic devices was recorded and analyzed to provide comparison data between filling time of a microfluidic chamber and surface modification characteristics, including the effects of surface energy and surface roughness on the microfluidic flow. We have experimentally demonstrated that fluid flow and thus filling time for the microfluidic device was significantly faster for the device with surface modifications that resulted in a lower static contact angle, and also that the incorporation of micro-pillars into a fluidic device increases the filling time when compared to comparative devices.

Proper bonding technique is investigated to achieve leakage-free surface-driven capillary flow in polymethylmethacrylate (PMMA) microfluidic devices. SU-8-based silicon stamp is fabricated by maskless lithography. This stamp is used to produce PMMA microchannel structure by hot embossing lithography. A direct bonding technique is mainly employed for leakage-free sealing inside PMMA microfluidic devices. The effect of surface wettability on surface-driven capillary flow is also investigated in PMMA microfluidic devices. The separation of polystyrene microparticles in PMMA laboratory-on-a-chip systems is investigated with the reduction of separation time by air dielectric barrier discharge (DBD) plasma processing of channel surfaces. This study is useful to fabricate the microfluidic laboratory-on-a-chip systems and to understand the surface-driven capillary flow.

Microchannel bends of rectangular cross-section were fabricated on polymethylmethacrylate (PMMA) by hot embossing, with a range of channel widths from 55 µm to 400 µm. The capillary movement of the interface between air and dyed water through the microchannels was recorded and analysed. The microfluidic flow behaviour as a function of the channel aspect ratio was studied. The evaluated Reynolds number was always less than 1.0 in each channel of every device. The air–water interface velocity in the devices was proportional to the channel aspect ratio. The air–water interface velocity shows a prominent increase at 90° separation angle. We observed an increasing trend of the air–water interface velocity with increasing channel aspect ratio. We have studied the separation of 10 µm polystyrene microparticles using a simple microchannel bend structure. We have obtained approximately 100% efficiency for the combined separation and clog-free blocking of 10 µm polystyrene microparticles using the above capillary flow behaviour in a modified microchannel bend structure.