Unpredictable Performance Research Articles

GAP: A group-based automatic pruning algorithm via convolution kernel fusion

In recent years, the deployment and operation of convolution neural networks on edge devices with limited computing capabilities have become increasingly challenging due to the large network structure and computational cost. Currently, the mainstream structured pruning algorithms mainly compress the network at the filter or layer level. However, these methods introduce too much human intervention with large granularities, which may lead to unpredictable performance after compression. In this paper, we propose a group-based automatic pruning algorithm(GAP) via kernel fusion to automatically search for the optimal pruning structure in a more fine-grained manner. Specifically, we first adopt a novel nonlinear dimensionality reduction clustering algorithm to divide the filters of each convolution layer into groups of equal size. Afterwards, we encode the mutual distribution similarity of the kernels within each group, and its KL divergence is employed as an importance indicator to determine the retained kernel groups through weighted fusion. Subsequently, we introduce an intelligent searching module that automatically explore and optimize the pruned structure of each layer. Finally, the pruned filters are permutated to form a dense group convolution and fine-tuned. Sufficient experiments show that, on two image classification datasets, for five advanced CNN models, our GAP algorithm outperforms most extant SOTA schemes, reduces artificial intervention, and enables efficient end-to-end training of compact models.

Nellix Device Failure Mechanisms Analysis on Explanted Grafts.

To understand possible reasons for poor durability of the Nellix (Endologix Inc., Irvine, USA) endovascular aneurysm sealing (EVAS) device. 21 Nellix endoprostheses explanted for endoleaks and migration underwent visual examinations of stent structures and instrumental examinations of the polymer endobags on 4 devices. We harvested 2.0-gram polymer slices out of each of them and tested the samples in an in vitro implantation replication that included wet and dry exposures. During the wet phase, we placed samples in a beaker with saline, mimicking the filling of the endobags during implantation. An exposure to a 37°C environment with 60% humidity during the dry phase replicated the postimplantation conditions inside the aneurysmal sac. Iatrogenic defects affected 16 (76%) metal stents and 20 (95%) endobags. The polymer was disintegrated owing to degradation in 15 (71%) cases. The polymer could lose more than 70% of its initial weight when partially dehydrated and regain 80% when placed in saline. We observed volume decrease and polymer fragmentation during these study phases. The polymer can lose weight and volume while it dehydrates. This structural degradation of the polymer could lead to the development of endoleaks and/or migration of the device. Based on the results of previous investigations, due to possible endovascular device degradation, patients with endografts should be offered life-long surveillance, and the Nellix device is no exception. Herein we suggest polymer degradation as one of the possible reasons for the device failure. Although Nellix has been withdrawn from the market, there are numerous patients with this type of endograft. Due to its unpredictable performance in the medium and long term, these patients should be recommended enhanced life-long surveillance every 6 months. Any suspicious conditions during the follow-up must be taken seriously and explantation should be considered.

Eraser: Eliminating Performance Regression on Learned Query Optimizer

Efficient query optimization is crucial for database management systems. Recently, machine learning models have been applied in query optimizers to generate better plans, but the unpredictable performance regressions prevent them from being truly applicable. To be more specific, while a learned query optimizer commonly outperforms the traditional query optimizer on average for a workload of queries, its performance regression seems inevitable for some queries due to model under-fitting and difficulty in generalization. In this paper, we propose a system called Eraser to resolve this problem. Eraser aims at eliminating performance regressions while still attaining considerable overall performance improvement. To this end, Eraser applies a two-stage strategy to estimate the model accuracy for each candidate plan, and helps the learned query optimizer select more reliable plans. The first stage serves as a coarse-grained filter that removes all highly risky plans with feature values that are seen for the first time. The second stage clusters plans in a more fine-grained manner and evaluates each cluster according to the prediction quality of learned query optimizers for selecting the final execution plan. Eraser can be deployed as a plugin on top of any learned query optimizer. We implement Eraser and demonstrate its superiority on PostgreSQL and Spark. In our experiments, Eraser eliminates most of the regressions while bringing very little negative impact on the overall performance of learned query optimizers, no matter whether they perform better or worse than the traditional query optimizer. Meanwhile, it is adaptive to dynamic settings and generally applicable to different database systems.

Numerical analysis of centrifugal water pump impeller under varying loads

Centrifugal pumps are widely used to move liquids from one place to another at varying pressure and flow rate, and to an elevated height for irrigation, water supply plants, and steam power plants. The load variations often lead to unpredictable performance of the impeller. The objective of this study is to investigate the performance of a three-stage centrifugal water pump impeller under varying loads using numerical methods. To conduct the investigation, a water pump used by Jimma town water treatment station, which operates with head of 191 m and discharges 439.2 m3/hr water was used as a case study. Computational fluid dynamics analysis of the pump was conducted using ANSYS 19.0 CFX to get the pressure distribution at different flow rates (60%, 100% and 140% of the design flow rate). The fluid pressure was then used to make static stress analysis of the impeller employing one way fluid structural interaction analysis for three different materials. The finding of the study shows that the fluid pressure on the impeller has great impact on the impeller performance causing high stress and deformation of the impeller.

Algorithm selection using edge ML and case-based reasoning

In practical data mining, a wide range of classification algorithms is employed for prediction tasks. However, selecting the best algorithm poses a challenging task for machine learning practitioners and experts, primarily due to the inherent variability in the characteristics of classification problems, referred to as datasets, and the unpredictable performance of these algorithms. Dataset characteristics are quantified in terms of meta-features, while classifier performance is evaluated using various performance metrics. The assessment of classifiers through empirical methods across multiple classification datasets, while considering multiple performance metrics, presents a computationally expensive and time-consuming obstacle in the pursuit of selecting the optimal algorithm. Furthermore, the scarcity of sufficient training data, denoted by dimensions representing the number of datasets and the feature space described by meta-feature perspectives, adds further complexity to the process of algorithm selection using classical machine learning methods. This research paper presents an integrated framework called eML-CBR that combines edge edge-ML and case-based reasoning methodologies to accurately address the algorithm selection problem. It adapts a multi-level, multi-view case-based reasoning methodology, considering data from diverse feature dimensions and the algorithms from multiple performance aspects, that distributes computations to both cloud edges and centralized nodes. On the edge, the first-level reasoning employs machine learning methods to recommend a family of classification algorithms, while at the second level, it recommends a list of the top-k algorithms within that family. This list is further refined by an algorithm conflict resolver module. The eML-CBR framework offers a suite of contributions, including integrated algorithm selection, multi-view meta-feature extraction, innovative performance criteria, improved algorithm recommendation, data scarcity mitigation through incremental learning, and an open-source CBR module, reshaping research paradigms. The CBR module, trained on 100 datasets and tested with 52 datasets using 9 decision tree algorithms, achieved an accuracy of 94% for correct classifier recommendations within the top k=3 algorithms, making it highly suitable for practical classification applications.

Stability analysis for traits related to seed yield in sunflower (Helianthus annuus L.)

An investigation was made to test the adaptability of twenty two hybrids along with four check hybrids for seedyield/plot(g), head diameter (cm) and hundred seed weight (g) over six environments. Environmental variances were significant for all the characterssuggesting the presence of genetic variability among genotypes over environments. The G x E (linear) were non-significant for all the characters under studied when tested against pooled deviation, indicates unpredictable performance of genotypes over the environments. The environment indicesrevealed that E2 was the most favourable environment for seed yield/plot (g), head diameter and 100-seed weight. For the stability parameter, none of the hybrid was ideal with an average adaptability for all the traits. The hybrid SVSH-473 had bi value nearer to unity (b=1), non-significant value of S2di and higher mean than the population mean indicating average stability and suitable for all environments. Based on the mean performance for head diameter the hybrid, SVSH-467 and for 100-seed weight the hybrid, SVSH-458 showed stable performance.

Insider selling and credit spread of corporate bonds

This paper investigates the impact of insider selling on credit spreads in the secondary market of corporate bonds. The data of listed corporate bonds with insider sale events from 2012 to 2021 are selected as the research sample. It is found that the greater the intensity of insider's holdings sale, the greater the credit spreads of corporate bonds in the secondary market; the mechanism test reveals that the more serious the company's unpredicted performance decline, the more pronounced the positive impact of insider's holdings sale on credit spreads; when analysts' attention increases, it correspondingly weakened the positive impact of insider's holdings sale on the credit spreads of corporate bonds.

CORRELATE OF INVENTORY MANAGEMENT AND ORGANIZATIONAL PERFORMANCE

CORRELATE OF INVENTORY MANAGEMENT AND ORGANIZATIONAL PERFORMANCE

Layered Design of a Highly Repeatable Electroactive Biofilm for a Standardized Biochemical Oxygen Demand Sensor.

Microbial electrochemical sensors are promising to monitor bioavailable organics in real environments, but their application is restricted by the unpredictable performance of the electroactive biofilm (EAB), which is randomly acclimated from environmental microflora. With a long-term stable EAB as a template, we successfully designed EAB (DEAB) by the sequential growth of Geobacter anodireducens and automatched microbes, achieving a reproducible high current than those naturally acclimated from wastewater (NEAB). Pre-inoculation of planktonic aerobes as oxygen bioscavengers was necessary to ensure the colonization of Geobacter in the inner layer, and the abundant Geobacter (50%) in DEAB guaranteed 4 times higher current density with a 15-fold smaller variation among 20 replicates than those of NEAB. The sensor constructed with DEAB exhibited a shorter measuring time and a precise biochemical oxygen demand (BOD) measurement with acetate, real domestic wastewater, and supernatant of anaerobic digestion. Here, we for the first time proposed an applicable strategy to standardize EABs for BOD sensors, which is also crucial to ensure a stable performance of all bioelectrochemical technologies.

GALAMC: Guaranteed Authentication Level at Minimized Complexity Relying on Intelligent Collaboration

Conventional centralized authentication techniques based on both digital cryptography and physical-layer attributes are prone to single-point failure due to either compromised digital security keys or an abrupt change in the physical communication environment. Although these particular challenges could be mitigated by the joint use of decentralized authentication and physical-layer attributes, such schemes often exhibit unpredictable performance. Simultaneously, the necessary involvement of multiple parties and the imperfect observation of the physical communication environment can also significantly increase the latency and computational complexity. As a remedy, a decentralized authentication scheme is proposed in this paper to achieve <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Guaranteed Authentication Level at Minimized Complexity</i> (GALAMC) based on the intelligent use of distributed collaboration and available distributive physical-layer attributes. Specifically, we aim for minimizing the complexity of the proposed collaborative authentication process by harnessing the minimum number of collaborative nodes and the selected authentication attributes at each node across the different environments while guaranteeing the required authentication level. The related physical-layer authentication scheme is implemented at each collaborative node where different physical-layer attributes can be selected based on their usefulness which is time-varying. The simulation results demonstrate that our scheme maintains the target level of authentication and it is more immune to sudden environmental changes than the conventional centralized physical-layer authentication scheme. It can also be observed that our proposed scheme can adaptively select the minimum number of collaborative nodes for adaptively minimizing the computational cost.

A relay velocity model infers cell-dependent RNA velocity

RNA velocity provides an approach for inferring cellular state transitions from single-cell RNA sequencing (scRNA-seq) data. Conventional RNA velocity models infer universal kinetics from all cells in an scRNA-seq experiment, resulting in unpredictable performance in experiments with multi-stage and/or multi-lineage transition of cell states where the assumption of the same kinetic rates for all cells no longer holds. Here we present cellDancer, a scalable deep neural network that locally infers velocity for each cell from its neighbors and then relays a series of local velocities to provide single-cell resolution inference of velocity kinetics. In the simulation benchmark, cellDancer shows robust performance in multiple kinetic regimes, high dropout ratio datasets and sparse datasets. We show that cellDancer overcomes the limitations of existing RNA velocity models in modeling erythroid maturation and hippocampus development. Moreover, cellDancer provides cell-specific predictions of transcription, splicing and degradation rates, which we identify as potential indicators of cell fate in the mouse pancreas.

In Operando Optical Tracking of Oxygen Vacancy Migration and Phase Change in few Nanometers Ferroelectric HZO Memories

AbstractFerroelectric materials offer a low‐energy, high‐speed alternative to conventional logic and memory circuitry. Hafnia‐based films have achieved single‐digit nm ferroelectricity, enabling further device miniaturization. However, they can exhibit nonideal behavior, specifically wake‐up and fatigue effects, leading to unpredictable performance variation over consecutive electronic switching cycles, preventing large‐scale commercialization. The origins are still under debate. Using plasmon‐enhanced spectroscopy, a non‐destructive technique sensitive to &lt;1% oxygen vacancy variation, phase changes, and single switching cycle resolution, the first real‐time in operando nanoscale direct tracking of oxygen vacancy migration in 5 nm hafnium zirconium oxide during a pre‐wake‐up stage is provided. It is shown that the pre‐wake‐up leads to a structural phase change from monoclinic to orthorhombic phase, which further determines the device wake‐up. Further migration of oxygen ions in the phase changed material is then observed, producing device fatigue. These results provide a comprehensive explanation for the wake‐up and fatigue with Raman, photoluminescence and darkfield spectroscopy, combined with density functional theory and finite‐difference time‐domain simulations.

CFIO: A conflict-free I/O mechanism to fully exploit internal parallelism for Open-Channel SSDs

I/O access conflicts make utilization within NVMe SSDs seriously low, which introduces unpredictable performance loss of NVMe SSDs. Although existing works adopt I/O isolation or conflict-aware I/O scheduling to avoid access conflicts, they can result in an unbalanced utilization and reduce the lifetime of NVMe SSDs. In this paper, we design and implement CFIO, a low-overhead conflict-aware I/O mechanism that achieves conflict-free I/Os to exploit the internal parallelism in NVMe SSDs. CFIO improves PU utilization and reduces I/O latency with two novel mechanisms. First, a conflict-free (CF) lane is proposed to eliminate conflicts by dividing I/O requests into conflict-free PU queues based on physical addresses. The PU queues correspond to the PU resources within the NVMe SSDs. Second, a k-RR scheduler is designed to dispatch reading and writing requests to NVMe SSDs in batches and separately. K-RR scheduler can fully exploit the internal parallelism of NVMe SSDs and form an I/O pipeline based on the dual registers of PU. Finally, we integrate CFIO into the LightNVM with Open-Channel NVMe SSD (OCSSD) and compare it with several existing solutions. Our evaluations show that CFIO improves the throughput of OCSSD by 19.32% and reduces its tail latency by 23.71%, compared to state-of-the-art methods.

TriangleKV: Reducing Write Stalls and Write Amplification in LSM-Tree Based KV Stores With Triangle Container in NVM

Popular LSM-tree based key-value stores suffer from suboptimal and unpredictable performance due to write amplification and write stalls that cause application performance to periodically drop to nearly zero. Our preliminary experimental studies reveal that (1) write stalls mainly stem from the significantly large amount of data involved in each compaction between <inline-formula><tex-math notation="LaTeX">$L_{0}$</tex-math></inline-formula> - <inline-formula><tex-math notation="LaTeX">$L_{1}$</tex-math></inline-formula> (i.e., the first two levels of LSM-tree), and (2) write amplification increases with the depth of LSM-trees. Existing work mainly focus on reducing write amplification, while only a couple of them target mitigating write stalls. In this paper, we exploit unique features of non-volatile memory (NVM) to address these two limitations and propose TriangleKV, a new LSM-tree based persistent KV store with multi-tier DRAM-NVM-SSD storage. TriangleKV's design principles include performing smaller and cheaper <inline-formula><tex-math notation="LaTeX">$L_{0}$</tex-math></inline-formula> - <inline-formula><tex-math notation="LaTeX">$L_{1}$</tex-math></inline-formula> compaction to reduce write stalls while reducing the depth of LSM-trees to mitigate write amplification. To this end, four novel techniques are proposed. First, we relocate and manage the <inline-formula><tex-math notation="LaTeX">$L_{0}$</tex-math></inline-formula> level in NVM with our proposed <i>triangle container</i> . Second, the new <i>right-angle side compaction</i> is devised to compact <inline-formula><tex-math notation="LaTeX">$L_{0}$</tex-math></inline-formula> to <inline-formula><tex-math notation="LaTeX">$L_{1}$</tex-math></inline-formula> at fine-grained key ranges, thus substantially reducing the amount of compaction data. Third, TriangleKV increases the width of each level to decrease the depth of LSM-trees thus mitigating write amplification. Finally, the <i>cross-row hint search</i> is introduced for the triangle container to keep adequate read performance. We implement TriangleKV based on MatrixKV and evaluate it on a hybrid DRAM/NVM/SSD system using Intel's latest 3D Xpoint NVM device Optane DC PMM. Evaluation results show that, with the same amount of NVM, TriangleKV outperforms RocksDB, NoveLSM and MatrixKV in 99th-percentile latencies by <inline-formula><tex-math notation="LaTeX">$5.5\times$</tex-math></inline-formula> , <inline-formula><tex-math notation="LaTeX">$2.1\times$</tex-math></inline-formula> and <inline-formula><tex-math notation="LaTeX">$1.1\times$</tex-math></inline-formula> , and random write throughput by <inline-formula><tex-math notation="LaTeX">$4.9\times$</tex-math></inline-formula> , <inline-formula><tex-math notation="LaTeX">$3.5\times$</tex-math></inline-formula> and <inline-formula><tex-math notation="LaTeX">$1.4\times$</tex-math></inline-formula> respectively.

IS MICRO-ENVIRONMENT A HURDLE IN GENETIC ENHANCEMENT OF YIELD IN PIGEONPEA?

Yield stagnation in pigeonpea remains a serious concern. Since most cultivars in the past were bread using pedigree method, it seems that this method was ineffective in increasing the crop productivity. To get insight into this failure, 10 highly inbred parents and their F2 bulks were examined by comparing their single plant yield variances. Surprisingly, the inbred parents exhibited high inter-plant variability; and their F2s were less variable than their one or both the parents. These unusual observations were attributed to the interaction of individual plants with unspecied micro-environmental factors prevailing in their vicinity, both below and above the ground. Under such circumstances, the selections had low heritability and unpredictable performance in the subsequent generation. Since these factors cannot be targeted through breeding, it is suggested that the breeders should try other breeding approaches such as bulk, composites, and single pod descent. The pedigree method of breeding, however, will continue to have its utility in the genetic enhancement of qualitative traits like yield.Yield stagnation in pigeonpea remains a serious concern. Since most cultivars in the past were bread using pedigree method, it seems that this method was ineffective in increasing the crop productivity. To get insight into this failure, 10 highly inbred parents and their F2 bulks were examined by comparing their single plant yield variances. Surprisingly, the inbred parents exhibited high inter-plant variability; and their F2s were less variable than their one or both the parents. These unusual observations were attributed to the interaction of individual plants with unspecied micro-environmental factors prevailing in their vicinity, both below and above the ground. Under such circumstances, the selections had low heritability and unpredictable performance in the subsequent generation. Since these factors cannot be targeted through breeding, it is suggested that the breeders should try other breeding approaches such as bulk, composites, and single pod descent. The pedigree method of breeding, however, will continue to have its utility in the genetic enhancement of qualitative traits like yield.

Overview of Big Data-Intensive Storage and its Technologies for Cloud and Fog Computing

Computing systems are becoming increasingly data-intensive because of the explosion of data and the needs for processing the data, and subsequently storage management is critical to application performance in such data-intensive computing systems. However, if existing resource management frameworks in these systems lack the support for storage management, this would cause unpredictable performance degradation when applications are under input/output (I/O) contention. Storage management of data-intensive systems is a challenge. Big Data plays a most major role in storage systems for data-intensive computing. This article deals with these difficulties along with discussion of High Performance Computing (HPC) systems, background for storage systems for data-intensive applications, storage patterns and storage mechanisms for Big Data, the Top 10 Cloud Storage Systems for data-intensive computing in today's world, and the interface between Big Data Intensive Storage and Cloud/Fog Computing. Big Data storage and its server statistics and usage distributions for the Top 500 Supercomputers in the world are also presented graphically and discussed as data-intensive storage components that can be interfaced with Fog-to-cloud interactions and enabling protocols.

Thermal Intelligent Control DVFS for Cyber-physical Systems

This work explores the use of well-known control techniques to tackle the heat dissipation problem in embedded devices. These types of devices lack of good solutions for managing heat and the existing ones may lead to unpredictable performance variations. We will show a way of solving the problem with classic control theory and workload identification.

Assessing Differential Item Functioning (DIF) For Pearson Test of English (PTE) a study of Test Takers with Different Fields of Study

Differential Item Functioning (DIF), which is a statistical feature of an item and provides a sign of unpredicted performance of items on a test, occurs once dissimilar groups of test takers with the same level of ability show different performance on a single test. The aim of this paper was to examine DIF on the Pearson Test of English (PTE) test items. To that end, 250 intermediate EFL learners with the age range of 26 - 36 in two different fields of study (125, Engineering, and 125 Sciences) were randomly chosen for the analysis. The Item Response Theory (IRT) Likelihood Ratio (LR) approach was utilized to find items showing DIF. The scored items of 250 PTE test takers were analyzed using the IRT three-parameter model utilizing item difficulty (b parameter), item discrimination (a parameter), and pseudo-guessing (c parameter). The results of the independent samples t-test for comparison of means in two groups depicted that Science participants performed better than the Engineering ones particularly in Speaking &amp; Writing and Reading sections. It is evident that the PTE test was statistically easier for the Science students at 0.05 level. Linguistic analyses of Differential Item Functioning items also confirmed the findings of the quantitative part, indicating a far better performance on the part of Science students.

Data-Driven Fault Diagnosis Techniques: Non-Linear Directional Residual vs. Machine-Learning-Based Methods.

Linear dependence of variables is a commonly used assumption in most diagnostic systems for which many robust methodologies have been developed over the years. In case the system nonlinearities are relevant, fault diagnosis methods, relying on the assumption of linearity, might potentially provide unsatisfactory results in terms of false alarms and missed detections. In recent years, many authors have proposed machine learning (ML) techniques to improve fault diagnosis performance to mitigate this problem. Although very powerful, these techniques require faulty data samples that are representative of any fault scenario. Additionally, ML techniques suffer from issues related to overfitting and unpredictable performance in regions which are not fully explored in the training phase. This paper proposes a non-linear additive model to characterize the non-linear redundancy relationships among the system signals. Using the multivariate adaptive regression splines (MARS) algorithm, these relationships are identified directly from the data. Next, the non-linear redundancy relationships are linearized to derive a local time-dependent fault signature matrix. The faulty sensor can then be isolated by measuring the angular distance between the column vectors of the fault signature matrix and the primary residual vector. A quantitative analysis of fault isolation and fault estimation performance is performed by exploiting real data from multiple flights of a semi-autonomous aircraft, thus allowing a detailed quantitative comparison with state-of-the-art machine-learning-based fault diagnosis algorithms.

Understanding the Effect of Interfacial Interphase on the Elastic Response of Hollow Glass Microsphere Reinforced Microcomposites

The hollow glass microspheres (HGMS) has been recently used in the fabrication of low-density polymeric composites due to rather high stiffness nature of the fillers together with their lightweight that in turn results in the development of micro-composites of engineered properties with enhanced mechanical properties. Interfacial interactions at the filler/polymer interface control the load transfer and, thus, bulk properties of composites leading to unpredictable performance of composites embedded with inclusions. Nevertheless, useful analytical models are required to estimate the mechanical behavior of the HGMS based composites with the incorporation of the effect of interfacial interactions and possible agglomeration of fillers. No studies so far have reported the analytical modeling of HGMS reinforced thermosetting composites emphasizing the role of the interphase shaped at the vicinity of fillers. This study aims at the fabrication of 0-20 wt% HGMS/polyester micro-composites followed by micromechanical modeling of the fabricated parts whilst the effect of the interphase region is emphasized by models modification. The results indicated a strong correlation between the interphase characteristics and Young’s modulus of the specimens revealing the dependency of the modulus on the thickness and modulus of the interphase as well as the level of agglomeration and interfacial debonding of the HGMSs. The results demonstrated that with considering no interphase, the models underestimate the modules of the parts, which suggests the presence of stiff interphase around the HGMS governed by changes in the interfacial cross-link density of the parent polymer as hypothesized supported by the mechanical response of the parts.