Missing Sample Research Articles

Low-rank representation induced missing-view recovery for incomplete multi-view clustering

In practical applications, multi-view data suffers from incompleteness, which refers to the absence of some samples in each view. Incomplete multi-view clustering is developed to capture the common cluster structure across incomplete views. However, some existing IMC methods ignore the missing samples or the latent relationship between the available and the unavailable samples. We propose a novel missing-view recovery strategy to overcome these two drawbacks. Considering that each missing sample and part of the observable samples belong to the same category, we recover each missing sample via the linear combination of observed samples within each view. Based on the semantic consistency, we extract a unified self-representation graph from complete views and impose a low-rank constraint on it. Furthermore, we learn a comprehensive representation from complete multi-view data and introduce the dynamic weights to consider the importance of various views. The experimental results on several datasets demonstrate that the proposed method outperforms the state-of-the-art IMC approaches.

Prediction of dissolved gas content in transformer oil based on multi‐information fusion

AbstractIn order to accurately predict the content and variation trend of dissolved gas in transformer oil and guide the condition maintenance of power transformers, a combined prediction model based on multi‐information fusion is proposed and its effectiveness is analysed. First of all, based on the possibility of pathological and missing historical sample data, a detection and filling method based on variable weight combination samples is established. Second, the authors propose two models. Aiming at the non‐linear and non‐stationary characteristics of gas content, a univariate decomposition prediction mode HBA‐VMD‐TCN which based on the Honey Badger algorithm, variational mode decomposition and time convolutional network (TCN) is established. Then the multivariate Informer prediction model is established for gas content affected by multiple variables. Third, the cross‐entropy theory is used to determine the weight coefficients of the two models, and the multi‐information fusion combined prediction model is formed. Finally, on the basis of the above, a method to determine the time step and the position information of the transition point adaptively in the process of prediction is proposed to further improve the prediction accuracy. The results show that, through a series of simulation experiments of model comparison and transformer anomaly prediction, the accuracy and effectiveness of the combined prediction model are verified.

An Introduction and Practical Guide to Strategies for Analyzing Longitudinal Data in Clinical Trials of Smoking Cessation Treatment: Beyond Dichotomous Point-Prevalence Outcomes.

Conceptualizing tobacco dependence as a chronic relapsing condition suggests the need to use analytic strategies that reflect that premise. However, clinical trials for smoking cessation typically define the primary endpoint as a measure of abstinence at a single timepoint distal to the intervention, typically 3-12 months. This reinforces the concept of tobacco outcomes as a dichotomous state-one is, or is not, abstinent. Fortunately, there are several approaches available to handle longitudinal data that reflect the relapsing and remitting nature of tobacco use during treatment studies. In this paper, sponsored by the Society for Research on Nicotine and Tobacco's Treatment Research Network, we present an introductory overview of these techniques and their application in smoking cessation clinical trials. Topics discussed include models to examine abstinence outcomes (eg, trajectory models of abstinence, models for transitions in smoking behavior, models for time to event), models that examine reductions in tobacco use, and models to examine joint outcomes (eg, examining changes in the use of more than one tobacco product). Finally, we discuss three additional relevant topics (ie, heterogeneity of effects, handling missing data, and power and sample size) and provide summary information about the type of model that can be used based on the type of data collected and the focus of the study. We encourage investigators to familiarize themselves with these techniques and use them in the analysis of data from clinical trials of smoking cessation treatment. Implications Clinical trials of tobacco dependence treatment typically measure abstinence 3-12 months after participant enrollment. However, because smoking is a chronic relapsing condition, these measures of intervention success may not accurately reflect the common trajectories of tobacco abstinence and relapse. Several analytical techniques facilitate this type of outcome modeling. This paper is meant to be an introduction to these concepts and techniques to the global nicotine and tobacco research community including which techniques can be used for different research questions with visual summaries of which types of models can be used for different types of data and research questions.

The Effectiveness of the StaySafe Intervention Using a Paradigm for Predicting Missing Outcome Data.

Sample attrition is a confounding issue in the analysis of data collected in follow-up studies. The present study uses a regression procedure that includes a propensity score as a predictor in estimating imputed data. The utility of the procedure was addressed by comparing results from this augmented data with those from the original data. Data were from a randomized controlled study testing the utility of a tablet-based intervention designed to improve decision-making with respect to health risk behaviors. Outcomes included self-reported testing for HIV, STD, and hepatitis. Two samples were used (163 in community facilities and 348 in residential facilities). Seventy-eight in the community sample and 238 in the residential sample completed follow-up surveys. Propensity scores based on a stepwise logistic regression were used to make the calibration sample and the missing data sample as close as possible. Multilevel analysis was performed for each outcome and multiple imputation compared estimated mean differences for the augmented and original analyses. The model imputing missing data was effective for the three outcomes and increased power. Least square mean differences between augmented and original data appeared to be essentially the same for most of the outcomes. This protocol has been registered with https://www.clinicaltrials.gov/(NCT02777086).

The MAGIC Algorithm Probability Predicts Long-Term Outcomes and Treatment Response to Second-Line Therapy for Acute Gvhd

Introduction: Risk stratification approaches using clinical criteria or biomarkers in acute graft-versus-host disease (GVHD) have traditionally been implemented at diagnosis or initiation systemic treatment with corticosteroids. For patients who require second-line treatment, the significance of biomarkers at salvage therapy and the relationship between biomarker risk and choice of therapy are unknown. Methods: We studied adult patients who received second-line treatment for acute GVHD between January 2016 and December 2021 in the Mount Sinai Acute GVHD Consortium (MAGIC) database, which collects clinical data and longitudinal serum samples from allogeneic HCT recipients using a rigorous PRoBE study design. Exclusion criteria included relapse of underlying malignant disease prior to second-line therapy, missing response data, or missing serum sample. Serum samples collected from 7 days prior to 3 days after initiation of second-line therapy met criteria for biomarker analysis. The MAGIC algorithm probability (MAP) was calculated as a single value based on concentrations of ST2 and REG3α and was stratified at a threshold of 0.290 to define high MAP and low MAP groups. The primary objective evaluated the association between MAP and survival for patients receiving second-line therapy for acute GVHD. The key secondary objective was to investigate the MAP and choice of second-line systemic agent on Day 28 overall response rate (ORR) and survival. Results: In total, 316 adult patients received second-line treatment for acute GVHD; 149 patients (47%) were excluded for disease relapse (n=27), missing response data (n=48), or missing sample (n=74).Of the 167 evaluable patients, most had steroid-refractory acute GVHD (n=146, 87%), rather than steroid-dependent (n=8, 5%) or steroid-sensitive (n=14, 8%) disease. 62 patients received ruxolitinib while 102 received other systemic agents (best available therapy: BAT). Baseline clinical characteristics of the study population according to treatment choice were similar, except for donor type (Rux: MRD 15%, MUD 74%, haplo 11%; BAT: MRD 30%, MUD 67%, haplo 3%; P=0.014) and median days from corticosteroids to second-line treatment (Rux: 18, BAT: 12; P=0.02).In total, 81 patients had high MAPs and 86 patients had low MAPs, with similar distribution in each treatment cohort (P=0.53). For the entire cohort, the median follow-up for survivors was 22 months (range, 1.5-27).The MAP separated patients into 2 groups (high vs low) with significantly different mortality (NRM at 2 years, 76% vs 29%, P<.0001) and survival (OS at 2 years, 12% vs 63%, P<.0001). Patients who received ruxolitinib as second-line therapy had lower mortality (NRM at 2 years, 35% vs 60%, P<.002) and improved survival (OS at 2 years, 51% vs 32%, P<.008), compared to those who received other systemic agents. However, the outcomes of patients with high MAP were poor, regardless of second-line treatment choice ( Figure 1A). The MAP (low vs high), time from corticosteroids to second-line treatment (≥14 days vs < 14 days), GVHD grade at second-line (0-2 vs 3-4), and choice of second-line therapy (Rux vs BAT) were all significant (P<0.05) predictors of lower NRM and higher OS in multivariate analysis. The MAP also separated patients into 2 groups with significantly different D28 ORR (low: 49%, high: 30%, P<.0001). In agreement with prospective trials, ruxolitinib resulted in higher D28 response rates compared to other systemic therapies (ORR 54.9% vs 30.5%, P=0.005), which was observed in both low MAP (ORR, 64.5% vs 38.5%, P=0.017) and high MAP cohorts (ORR, 42.9% vs 22.6%, P=0.06) ( Figure 1B). The MAP (low vs high), GVHD grade at second line (0-2 vs 3-4), and choice of second-line therapy (Rux vs BAT) were significant predictors (P<0.05) of higher D28 ORR in multivariate analysis. Conclusion: The MAP measured at the initiation of second-line systemic treatment for acute GVHD predicts NRM and OS, as well as treatment response. The outcomes of patients with high MAP are poor, regardless of second-line treatment choice, and the higher CR rate in low MAP patients may drive higher survival rates. Incorporation of MAP into clinical trials in this setting warrants investigation. The results also support the current use of ruxolitinib as second-line treatment for acute GVHD, particularly in patients with low MAP.

Immunogenicity of bivalent omicron (BA.1) booster vaccination after different priming regimens in health-care workers in the Netherlands (SWITCH ON): results from the direct boost group of an open-label, multicentre, randomised controlled trial

Bivalent mRNA-based COVID-19 vaccines encoding the ancestral and omicron spike (S) protein were developed as a countermeasure against antigenically distinct SARS-CoV-2 variants. We aimed to assess the (variant-specific) immunogenicity and reactogenicity of mRNA-based bivalent omicron (BA.1) vaccines in individuals who were primed with adenovirus-based or mRNA-based vaccines encoding the ancestral spike protein. We analysed results of the direct boost group of the SWITCH ON study, an open-label, multicentre, randomised controlled trial. Health-care workers from four academic hospitals in the Netherlands aged 18-65 years who had completed a primary COVID-19 vaccination regimen and received one booster of an mRNA-based vaccine, given no later than 3 months previously, were eligible. Participants were randomly assigned (1:1) using computer software in block sizes of 16 and 24 to receive an omicron BA.1 bivalent booster straight away (direct boost group) or a bivalent omicron BA.5 booster, postponed for 90 days (postponed boost group), stratified by priming regimen. The BNT162b2 OMI BA.1 boost was given to participants younger than 45 years, and the mRNA-1273.214 boost was given to participants 45 years or older, as per Dutch guidelines. The direct boost group, whose results are presented here, were divided into four subgroups for analysis: (1) Ad26.COV2.S (Johnson & Johnson) prime and BNT162b2 OMI BA.1 (BioNTech-Pfizer) boost (Ad/P), (2) mRNA-based prime and BNT162b2 OMI BA.1 boost (mRNA/P), (3) Ad26.COV2.S prime and mRNA-1273.214 (Moderna) boost (Ad/M), and (4) mRNA-based prime and mRNA-1273.214 boost (mRNA/M). The primary outcome was fold change in S protein S1 subunit-specific IgG antibodies before and 28 days after booster vaccination. The primary outcome and safety were assessed in all participants except those who withdrew, had a SARS-CoV-2 breakthrough infection, or had a missing blood sample at day 0 or day 28. This trial is registered with ClinicalTrials.gov, NCT05471440. Between Sept 2 and Oct 4, 2022, 219 (50%) of 434 eligible participants were randomly assigned to the direct boost group; 187 participants were included in the primary analyses; exclusions were mainly due to SARS-CoV-2 infection between days 0 and 28. From the 187 included participants, 138 (74%) were female and 49 (26%) were male. 42 (22%) of 187 participants received Ad/P and 44 (24%) mRNA/P (those aged <45 years), and 45 (24%) had received Ad/M and 56 (30%) mRNA/M (those aged ≥45 years). S1-specific binding antibody concentrations increased 7 days after bivalent booster vaccination and remained stable over 28 days in all four subgroups (geometric mean ratio [GMR] between day 0 and day 28 was 1·15 [95% CI 1·12-1·19] for the Ad/P group, 1·17 [1·14-1·20] for the mRNA/P group, 1·20 [1·17-1·23] for the Ad/M group, and 1·16 [1·13-1·19] for the mRNA/M group). We observed no significant difference in the GMR between the Ad/P and mRNA/P groups (p=0·51). The GMR appeared to be higher in the Ad/M group than in the mRNA/M group, but was not significant (p=0·073). Most side-effects were mild to moderate in severity and resolved within 48 h in most individuals. Booster vaccination with mRNA-1273.214 or BNT162b2 OMI BA.1 in adult healthcare workers resulted in a rapid recall of humoral and cellular immune responses independent of the priming regimen. Monitoring of SARS-CoV-2 immunity at the population level, and simultaneously antigenic drift at the virus level, remains crucial to assess the necessity and timing of COVID-19 variant-specific booster vaccinations. The Netherlands Organization for Health Research and Development (ZonMw).

Investigating on the robustness of flow-based intrusion detection system against adversarial samples using Generative Adversarial Networks

Recently, Software Defined Networking (SDN) has emerged as the key technology in programming and orchestrating security policy in the security operations centers (SOCs) for heterogeneous networks. Typically, machine learning-based intrusion detection systems (ML-IDS) have been deployed and associated with SDN to leverage the features of a programmable network to defend against sophisticated cyberattacks in anomaly detection. Unfortunately, such ML-based IDSs are easily vulnerable to adversarial attacks due to the lack of diverse forms of malicious records in the training dataset. The missing data sample in the training phase can lead to a lower detection rate in real-world scenarios with adversarial settings. In this paper, we explore the ability of Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN-GP), WGAN-GP with two timescale update rule (WGAN-GP TTUR), and AdvGAN in generating perturbed attack samples to bypass attack detectors. Then, this approach is used to continuously evaluate the robustness of ML-based IDSs and then upgrade them as a service in SDN. The experimental results on CICIDS2018 and InSDN datasets demonstrate that generated adversarial samples can be used to fool targeted IDS. Later, those created samples can supplement the original ones in retraining IDS to improve the resilience of the attack detector.

EFCMF: A Multimodal Robustness Enhancement Framework for Fine-Grained Recognition

Fine-grained recognition has many applications in many fields and aims to identify targets from subcategories. This is a highly challenging task due to the minor differences between subcategories. Both modal missing and adversarial sample attacks are easily encountered in fine-grained recognition tasks based on multimodal data. These situations can easily lead to the model needing to be fixed. An Enhanced Framework for the Complementarity of Multimodal Features (EFCMF) is proposed in this study to solve this problem. The model’s learning of multimodal data complementarity is enhanced by randomly deactivating modal features in the constructed multimodal fine-grained recognition model. The results show that the model gains the ability to handle modal missing without additional training of the model and can achieve 91.14% and 99.31% accuracy on Birds and Flowers datasets. The average accuracy of EFCMF on the two datasets is 52.85%, which is 27.13% higher than that of Bi-modal PMA when facing four adversarial example attacks, namely FGSM, BIM, PGD and C&amp;W. In the face of missing modal cases, the average accuracy of EFCMF is 76.33% on both datasets respectively, which is 32.63% higher than that of Bi-modal PMA. Compared with existing methods, EFCMF is robust in the face of modal missing and adversarial example attacks in multimodal fine-grained recognition tasks. The source code is available at https://github.com/RPZ97/EFCMF (accessed on 8 January 2023).

MRD4 Eradication at 6 Months and Early Clearance of MRD with Combination of Ibrutinib Plus Venetoclax Results in Sustained Clinical and MRD Responses: Exploratory Analysis of the Blood Cancer UK TAP Clarity Study

Background: Treatment of chronic lymphocytic leukemia (CLL) has been revolutionized by targeting the B-cell receptor pathway with inhibitors such as ibrutinib (IBR), and molecules that reactivate apoptosis such as venetoclax (VEN). The Blood Cancer UK TAP CLARITY trial combined IBR with VEN to eradicate detectable CLL with the intention of stopping therapy. Methods: CLARITY is a Phase II trial combining IBR with VEN in 50 patients with relapsed/refractory (R/R) CLL. After 2 months of IBR monotherapy, VEN was added at a daily dose of 10mg/day or 20mg/day, escalating weekly to a final daily dose of 400mg/day. The primary outcome was the proportion of patients eradicating MRD (<0.01%) in the bone marrow (BM) at Month 14. Secondary outcomes included the proportion of patients eradicating MRD at Months 8, 26 and 38 and IWCLL response at Months 14, 26 and 38, progression-free and overall survival. CLL MRD was quantified using >6 color ERIC-standard flow cytometry (detection limit 10-5/0.001%). Paired PB and BM samples were assessed at Months 8, 14, 26 and longitudinal PB samples were taken at additional intermediate time points. Therapy was discontinued upon confirmation of a CR according to the IWCLL2008 criteria and MRD reduction to <10-4 CLL cells in PB and BM (MRD4). MRD4 at Month 8 indicated to stop IBR and VEN at Month 14; MRD4 in PB and BM at Months 14 and/or 26 indicated to stop IBR and VEN at Month 26; while detectable MRD at Month 26 indicated to stop VEN and to continue IBR until progression or toxicity. The original trial protocol limited venetoclax to 24 months of treatment, but this was later amended in January 2019 to allow patients, who had not obtained an MRD4 at month 26, to have an extra 12 months of VEN in addition to IBR. The aim of this exploratory analysis is to ascertain whether 2 log reduction in MRD at 2 months of combination or achievement of MRD4 at 6 months of combination results in persistence of MRD and clinical responses with prolonged follow-up. Results: Fifty-four patients were recruited from May 2016 to November 2017. The median number of prior therapies was 1 (range: 1-6). 20% patients had del(17p); 25% del(11q); 75% had unmutated IGVH. Four patients that discontinued treatment due to ibrutinib-related adverse events in the first 8 weeks before starting VEN were replaced so that, in total, 50 patients received the combination of IBR and VEN. Of the 50 evaluable participants recruited to the study, 14 patients achieved MRD-negativity at a level of <0.01% in PB and BM at 6 months of combination of IBR and VEN. At month 38 of follow-up, one patient has relapsed but not needed treatment. 10/14 patients continue at MRD4 levels and 3/14 have a missing MRD sample at Month 38 but all three patients were MRD-negative at Month 26. There were no clinical progression in the 13 patients to date. Of the 25 patients achieving 2 log reduction in MRD levels at 2 months of combination of IBR and VEN, 15/25 patients were still MRD-negative at levels of <0.01% at month 38; 4/25 patients were MRD-positive at Month 38. 6/25 patients had missing samples at month 38; 5 patients were still MRD-negative at month 26 with only one patient being MRD-positive at month 26. There were two clinical progressions to date at month 41 and 42 respectively. Conclusions: Early MRD eradication with IBR+VEN results in sustained MRD and clinical responses in R/R CLL. Achievement of MRD <0.01% with 6 months of combination or 2 log reduction in MRD levels at 2 months of combination co-related with sustained MRD and clinical responses at 3 years follow-up suggesting that this approach may help to ascertain the best responders to this combination.

Bias dynamics for parameter estimation with missing data mechanisms under logistic model

Missing data affects the validity of statistical conclusions through biasing estimated parameters, increasing standard errors and reducing statistical power. We evaluated the effect of missingness, sample size and missing data mechanisms on bias of the estimated parameters. Our findings, based on survey and simulated data indicate that higher proportions of missing data and smaller sample size considerably increase bias in estimated parameters. Further, bias was higher for the missing at random than missing completely at random mechanism. Moreover, whereas multiple imputation renders a viable solution to the missing data problem, imputation with more than 35% of missing data may generate unreliable model estimates.

Bootstrap joint prediction regions for sequences of missing values in spatio-temporal datasets

Missing data reconstruction is a critical step in the analysis and mining of spatio-temporal data. However, few studies comprehensively consider missing data patterns, sample selection and spatio-temporal relationships. To take into account the uncertainty in the point forecast, some prediction intervals may be of interest. In particular, for (possibly long) missing sequences of consecutive time points, joint prediction regions are desirable. In this paper we propose a bootstrap resampling scheme to construct joint prediction regions that approximately contain missing paths of a time components in a spatio-temporal framework, with global probability 1-alpha . In many applications, considering the coverage of the whole missing sample-path might appear too restrictive. To perceive more informative inference, we also derive smaller joint prediction regions that only contain all elements of missing paths up to a small number k of them with probability 1-alpha . A simulation experiment is performed to validate the empirical performance of the proposed joint bootstrap prediction and to compare it with some alternative procedures based on a simple nominal coverage correction, loosely inspired by the Bonferroni approach, which are expected to work well standard scenarios.

Three‐step imputation of missing values in condition monitoring datasets

Missing values are a common occurrence in condition monitoring datasets. To effectively improve the integrity of data, many data imputation methods have been developed to replace the missing values with the estimated values. However, these methods do not always perform well in datasets containing different types of missing values. Three types of missing data are defined, namely isolated missing value, continuous missing variable, and continuous missing sample. A three-step data imputation method is proposed to sequentially impute these missing values following the principle from easy to difficult. The original time series data is first to split into different segments according to the positions of continuous missing samples. Then, interpolation and space-based methods are applied to sequentially estimate isolated missing values and continuous missing variables in each segment. Finally, a stepwise extrapolation prediction model based on the long short-term memory network is established to repair continuous missing samples between each segment. Two application examples are implemented on different dissolved gas analysis datasets and load datasets. Compared with state-of-the-art techniques, the proposed three-step data imputation method is general and can be applied to many scenarios because it establishes a rational data recovery sequence to accurately repair both stationary and non-stationary condition monitoring data.

Low-Rank Hankel Matrix Completion for Robust Time-Frequency Analysis

In this paper, we develop a novel method to enable robust sparsity-based time-frequency representation of multi-component frequency modulated signals in the presence of burst missing samples, where the amplitudes of the different signal components are generally different. Unlike existing methods which require cross-term presence to be sparse in the time-frequency domain, the proposed method permits effective time-frequency representation reconstruction even when undesired cross-terms take a high occupancy. A key enabling procedure is the high-fidelity missing entry recovery of the instantaneous autocorrelation function that is insensitive to cross-terms. By designing instantaneous autocorrelation function patches such that their Doppler-frequency domain representation is sparse, we formulate the instantaneous autocorrelation function recovery problem as a patch-based low-rank block Hankel matrix completion problem. This approach effectively suppresses the effects of burst missing data samples and is robust to the amplitude differences. A data-adaptive time-frequency kernel is then applied to further mitigate the undesired cross-terms and the residual artifacts due to the burst missing samples. We prove the superiority of the proposed method over the state of the art for both multi-component linear and nonlinear frequency modulated signals. Simulation results confirm that the proposed method outperforms the state of the art for different types of frequency modulated signals with varying signal-to-noise ratios and missing sample rates.

Sparse recovery of missing image samples using a convex similarity index

Abstract This paper investigates the problem of recovering missing samples using methods based on sparse representation adapted for visually enhanced quality of reconstruction of image signals. Although, the popular Mean Square Error (MSE) criterion is convex and simple, it may not be entirely consistent with Human Visual System (HVS). Thus, instead of l2-norm or MSE, a new perceptual quality measure is used as the similarity criterion between the original and the reconstructed images. The proposed criterion called Convex SIMilarity (CSIM) index is a modified version of the Structural SIMilarity (SSIM) index, which despite its predecessor, is convex and uni-modal. We derive mathematical properties for the proposed index and show how to optimally choose the parameters of the proposed criterion, investigating the Restricted Isometry (RIP) and error-sensitivity properties. We also propose an iterative sparse recovery method based on a constrained l1-norm minimization problem, incorporating CSIM as the fidelity criterion. The resulting convex optimization problem is solved via an algorithm based on Alternating Direction Method of Multipliers (ADMM). Taking advantage of the convexity of the CSIM index, we also prove the convergence of the algorithm to the globally optimal solution of the proposed optimization problem, starting from any arbitrary point. Simulation results confirm the performance of the new similarity index as well as the proposed algorithm for missing sample recovery of image patch signals.

Partial Multi-View Outlier Detection Based on Collective Learning

In the past decade, various multi-view outlier detection methods have been designed to detect horizontal outliers that exhibit inconsistent across-view characteristics. The existing works assume that all objects are present in all views. However, in real-world applications, it is often the incomplete case that every view may suffer from some missing samples, resulting in partial objects difficult to detect outliers from. To address this problem, we propose a novel Collective Learning (CL) based framework to detect outliers from partial multi-view data in a self-guided way. More specifically, by well exploiting the inter-dependence among different views, we develop an algorithm to reconstruct missing samples based on learning. Furthermore, we propose similarity-based outlier detection to break through the dilemma that the number of clusters is unknown priori. Then, the calculated outlier scores act as the confidence levels in CL and in turn guide the reconstruction of missing data. Learning-based missing sample recovery and similarity-based outlier detection are iteratively performed in a self-guided manner. Experimental results on benchmark datasets show that our proposed approach consistently and significantly outperforms state-of-the-art baselines.

Bayesian hypothesis testing detector for one bit diffusion LMS with blind missing samples

This paper proposes a sparse distributed estimation algorithm when missing data occurs in the measurements over adaptive networks. Two classes of measurement models are considered. First, the traditional linear regression model is investigated and second the sign of the linear regression model is studied. The latter is referred to as one-bit model. We utilize the diffusion LMS strategy, in the proposed methods, where a set of nodes cooperates with each other to estimate a vector model parameter. In both models, it is shown that replacing the missing sample with a simple estimate is equivalent to removing the missing sample from the distributed diffusion algorithm. We consider two cases, where in the first case the positions of missing samples are known (non-blind) and in the second case the positions of missing samples are unknown (blind). In the linear regression model scenario, a Bayesian hypothesis testing (BHT) is used for detection of the missing samples. In the one-bit model, in addition to BHT detector, a simple heuristic detector, based on mean square error (MSE), is also suggested. Simulation results show the effectiveness of the proposed detector-assisted distributed algorithms.

A Two-Step Method for Missing Spatio-Temporal Data Reconstruction

Missing data reconstruction is a critical step in the analysis and mining of spatio-temporal data; however, few studies comprehensively consider missing data patterns, sample selection and spatio-temporal relationships. As a result, traditional methods often fail to obtain satisfactory accuracy or address high levels of complexity. To combat these problems, this study developed an effective two-step method for spatio-temporal missing data reconstruction (ST-2SMR). This approach includes a coarse-grained interpolation method for considering missing patterns, which can successfully eliminate the influence of continuous missing data on the overall results. Based on the results of coarse-grained interpolation, a dynamic sliding window selection algorithm was implemented to determine the most relevant sample data for fine-grained interpolation, considering both spatial and temporal heterogeneity. Finally, spatio-temporal interpolation results were integrated by using a neural network model. We validated our approach using Beijing air quality data and found that the proposed method outperforms existing solutions in term of estimation accuracy and reconstruction rate.

A novel dictionary-based approach for missing sample recovery of signals in manifold

Recovery of signal with randomly positioned missing samples is a difficult or impossible task. The aim of this paper is to accurately recover missing samples if prior information on domain of sparsity is known. This paper proposes a novel approach for recovery of signals lying in low-dimensional sub-manifold, embedded in high-dimensional signal space and heavily corrupted by arbitrarily positioned missing samples. The proposed simple and efficient algorithm is based on global manifold model and can recover the corrupted signal from the limited available samples without affecting the remaining samples. The proposed method is applicable to any type of data and can be extensively used in a wide variety of data processing techniques. Experimental results prove that the proposed method outperforms the counterparts without much computational complexity.

Analysis of Longitudinal Outcome Data with Missing Values in Total Knee Arthroplasty

We sought to determine the influence of missing data on the statistical results, and to determine which statistical method is most appropriate for the analysis of longitudinal outcome data of TKA with missing values among repeated measures ANOVA, generalized estimating equation (GEE) and mixed effects model repeated measures (MMRM). Data sets with missing values were generated with different proportion of missing data, sample size and missing-data generation mechanism. Each data set was analyzed with three statistical methods. The influence of missing data was greater with higher proportion of missing data and smaller sample size. MMRM tended to show least changes in the statistics. When missing values were generated by ‘missing not at random’ mechanism, no statistical methods could fully avoid deviations in the results.

A Hybrid Algorithm for Online Location Update using Feature Point Detection for Portable Devices

We propose a cost-efficient hybrid algorithm for online location updates that efficiently combines feature point detection with the online trajectory-based sampling algorithm. Our algorithm is designed to minimize the average trajectory error with the minimal number of sample points. The algorithm is composed of 3 steps. First, we choose corner points from the map as sample points because they will most likely cause fewer trajectory errors. By employing the online trajectory sampling algorithm as the second step, our algorithm detects several missing and important sample points to prevent unwanted trajectory errors. The final step improves cost efficiency by eliminating redundant sample points on straight paths. We evaluate the proposed algorithm with real GPS trajectory data for various bus routes and compare our algorithm with the existing one. Simulation results show that our algorithm decreases the average trajectory error 28% compared to the existing one. In terms of cost efficiency, simulation results show that our algorithm is 29% more cost efficient than the existing one with real GPS trajectory data.