Ambiguous Prediction Research Articles

Random projection ensemble conformal prediction for high-dimensional classification

In classification problems, many models with superior performance fail to provide confidence estimates or intervals for each prediction. This lack of reliability poses risks in real-world applications, making these models difficult to trust. Conformal prediction, as distribution-free and model-free approaches with finite-sample coverage guarantee, have recently been widely used to construct prediction sets for classification models. However, traditional conformal prediction methods only produce set-valued results without specifying a definitive predicted class. Particularly in complex settings, these methods fail to assist models in effectively addressing challenges such as high dimensionality, resulting in ambiguous prediction sets with low statistical efficiency, i.e. the prediction sets contain many false classes. In this study, a novel Ensemble Conformal Prediction algorithm based on Random Projection and a designed voting strategy, RPECP, is developed to tackle these challenges. Initially, a procedure for selecting the approximately oracle random projections and classifiers is executed to best leverage the internal information and structure of the data. Subsequently, based on the approximately oracle random projections and underlying classifiers, conformal prediction is performed on new test samples in a lower-dimensional space, resulting in multiple independent prediction sets. Finally, an accurate predicted class and a precise prediction set with high coverage and statistical efficiency are produced through a designed voting strategy. Compared to several base classifiers, RPECP obtain higher classification accuracy; against other conformal prediction algorithms, it achieves less ambiguous prediction sets with fewer false classes while guaranteeing high coverage. For illustration, this paper demonstrates RPECP's superiority over other methods in four cases: two high-dimensional settings and two real-world datasets.

Risk preferences, preventive behaviour, and the probability of a loss: Empirical evidence from the COVID-19 pandemic

A theoretical model of optimal choice under risk, in which an individual chooses the level of prevention to avoid a loss, has the ambiguous prediction that a higher risk-taking preference increases the probability of a loss. To empirically investigate the prediction in the case of COVID-19 with individual-level survey data. Survey data from the Understanding America Study (UAS). The UAS Coronavirus Tracking Survey followed 8628 respondents from March 2020 until July 2021 (29 survey waves) and data was gathered on having contracted COVID-19, vaccination, and preventive behaviour. Separate UAS modules gathered data on individuals' risk preferences; twice before and once during the COVID-19 pandemic. UAS also gathered data on pre-pandemic health and socio-economic status. Combining these data, and dropping missing observations, provided longitudinal data for 4335 respondents (96,370 observations) of whom 530 contracted COVID-19. In support of the theoretical prediction, the empirical findings show that a one-standard deviation higher risk-taking preference is associated with about a one-third higher probability of contracting COVID-19 within two weeks. Furthermore, the findings show that individuals' risk-taking preference is negatively associated with the preventive behaviour of social distancing and not associated with getting vaccinated. There is, however, no support for preventive behaviour being associated with the probability of contracting COVID-19. The exception is for being vaccinated, which is negatively associated with the probability of contracting COVID-19. The findings, therefore, do not support that the positive association of the risk-taking preference with the probability of contracting COVID-19 is mediated through observed preventive behaviour. The findings support the importance of individuals' risk-taking behaviour for contracting COVID-19 and, more generally, the importance of loss prevention as a risk management tool for individuals.

Improving Compound-Protein Interaction Prediction by Self-Training with Augmenting Negative Samples.

Identifying compound-protein interactions (CPIs) is crucial for drug discovery. Since experimentally validating CPIs is often time-consuming and costly, computational approaches are expected to facilitate the process. Rapid growths of available CPI databases have accelerated the development of many machine-learning methods for CPI predictions. However, their performance, particularly their generalizability against external data, often suffers from a data imbalance attributed to the lack of experimentally validated inactive (negative) samples. In this study, we developed a self-training method for augmenting both credible and informative negative samples to improve the performance of models impaired by data imbalances. The constructed model demonstrated higher performance than those constructed with other conventional methods for solving data imbalances, and the improvement was prominent for external datasets. Moreover, examination of the prediction score thresholds for pseudo-labeling during self-training revealed that augmenting the samples with ambiguous prediction scores is beneficial for constructing a model with high generalizability. The present study provides guidelines for improving CPI predictions on real-world data, thus facilitating drug discovery.

Investigating and Comparing the Applicability of the R3m Molecular Descriptor and Solubility Parameter Estimation Approaches in Predicting Dispersion Formation Potential of APIs in a Random Co-Polymer Polyvinylpyrrolidone Vinyl Acetate and its Homopolymer

Evaluation of different amorphous solid dispersion carrier matrices is enabled by active pharmaceutical ingredient (API) structure-based predictions. This study compares the utility of Hansen Solubility Parameters with the R3m molecular descriptor for identifying dispersion polymers based on the structure of the drug molecule. Twelve API-polymer combinations (4 APIs and 3 interrelated polymers) were used to test each approach. Co-solidified mixtures containing 75% API were prepared by melt-quenching. Phase behavior was evaluated and classified using differential scanning calorimetry, powder X-ray diffraction, polarized light microscopy, and hot stage microscopy. Observations of dispersion behavior were compared to predictions made using the Hansen Solubility Parameter and R3m. The solubility parameter approach misclassified the dispersion behavior of 1 API-polymer combination and also did not produce definite predictions in 3 out of 12 of the API-polymer combinations. In contrast, R3m classifications of dispersion behavior were correct in all but two cases, with one misclassification and one ambiguous prediction. The solubility parameters best classify dispersion behavior when specific drug-polymer intermolecular interactions are present, but may be less useful otherwise. Ultimately, these two methods are most effectively used together, as they are based on distinct features of the same molecular structure.

A comparative study on estimation of fractal dimension of EMG signal using SWT and FLP

ABSTRACT Most of the biomedical signals gathered from various electronic devices contain a certain amount of noise based on the external atmosphere, the working condition of the devices and the activity of the clinician handling the observation section. Analysing such signals may result in an ambiguous prediction of the patients’ diagnosis. Thus, the eradication of noises in the biosignal waves gets a tremendous attraction in the medical field. Hence, a specific aim of this study is to handle two perfect and precise methods of clearing the defectiveness in the electromyography (EMG) signals which are termed stationary wavelet transform (SWT) and fractional linear prediction (FLP). Once we get the filtered signal from these two techniques, we can calculate the fractal dimension of the denoised EMG signals for obtaining the accurate way for offering medication. In this regard, the present study primarily reveals the comparison between fractal dimensions that are gained through the following specified methods, Higuchi, Katz, rescaled-range analysis, relative dispersion and power spectral density. Ultimately, these dimensions are compared to show a clear-cut result by considering both the filtering mechanisms. For the comparison of statistical data of an undertaken sample, mathematics provides an estimating procedure called the ANOVA test. This paper also deals with some of the statistical measures for comparing the denoising methods and results to attain a better one. As a consequence, we have concluded that the FLP filtering method is more reliable than the SWT technique, based on the fractal dimension forecasted via the five mechanisms.

Modified naive Bayes classifier for classification of protein-protein interaction sites

The prediction of protein-protein interaction sites (PPIs) is a vital importance in biology for understanding the physical and functional interactions between molecules in living systems. There are several classification approaches for the prediction of PPI sites; the naïve Bayes classifier is one of the most popular candidates. But the ordinary naïve Bayes classifier is sensitive to unusual protein sequence profiling feature dataset and sometimes it gives ambiguous prediction results. To overcome this problem we have been modified the naïve Bayes classifier by radial basis function (RBF) kernel for the prediction of PPI sites. We investigate the performance of our proposed method compared with the popular classifiers like linear discriminant analysis (LDA), naïve Bayes classifier (NBC), support vector machine (SVM), AdaBoost and k-nearest neighbor (KNN) by the protein sequence profiling data analysis. The mNBC method showed sensitivity (86%), specificity (81%), accuracy (83%) and MCC (65%) for prediction of PPI sites.

Space Weather Monitor at the L5 Point: A Case Study of a CME Observed with STEREO B

AbstractAn important location for future space weather monitoring is the Lagrange point 5 (L5) of the Sun‐Earth system. We test the performance of L5 for space weather monitoring using STEREO B observations of an Earth‐directed coronal mass ejection (CME), seen as a partial halo by SOHO at L1. STEREO B (located close to L5) continuously tracked the CME. By using these data in combination with methods to calculate the CME arrival time at the Earth (extrapolation, drag‐based model, and a magnetohydrodynamic model), we demonstrate that the estimation of the CME arrival time can be drastically improved by adding L5 data. Based on the L1 data alone, one could predict that the CME would arrive at the Earth. Using only the L5 data, one would not expect an arrival, as the estimations of the CME 3‐D configuration is uncertain. The combination of L1 and L5 data leads to an ambiguous prediction of the CME arrival due to low CME brightness in L1 data. To obtain an unambiguous prediction, one needs its 3‐D configuration, from observing the CME material close to the plane of the sky from at least two viewpoints (in this case L5 and, coincidentally, L4). This event demonstrates that L1 observations may be better to determine CME arrival, but L5 observations are superior for constraining arrival time. In this work, the advantages and caveats of using data from a space weather monitor at L5 for predicting interplanetary propagation of CMEs are discussed and demonstrated in a direct case study.

Road Extraction by Using Atrous Spatial Pyramid Pooling Integrated Encoder-Decoder Network and Structural Similarity Loss

The technology used for road extraction from remote sensing images plays an important role in urban planning, traffic management, navigation, and other geographic applications. Although deep learning methods have greatly enhanced the development of road extractions in recent years, this technology is still in its infancy. Because the characteristics of road targets are complex, the accuracy of road extractions is still limited. In addition, the ambiguous prediction of semantic segmentation methods also makes the road extraction result blurry. In this study, we improved the performance of the road extraction network by integrating atrous spatial pyramid pooling (ASPP) with an Encoder-Decoder network. The proposed approach takes advantage of ASPP’s ability to extract multiscale features and the Encoder-Decoder network’s ability to extract detailed features. Therefore, it can achieve accurate and detailed road extraction results. For the first time, we utilized the structural similarity (SSIM) as a loss function for road extraction. Therefore, the ambiguous predictions in the extraction results can be removed, and the image quality of the extracted roads can be improved. The experimental results using the Massachusetts Road dataset show that our method achieves an F1-score of 83.5% and an SSIM of 0.893. Compared with the normal U-net, our method improves the F1-score by 2.6% and the SSIM by 0.18. Therefore, it is demonstrated that the proposed approach can extract roads from remote sensing images more effectively and clearly than the other compared methods.

Statins in conditions other than hypocholesterolemic effects for chronic subdural hematoma therapy, old drug, new tricks?

Chronic subdural hematoma (CSDH) is one of the most common intracranial hematomas worldwide with a high incidence in the general population. However, the optimum treatment for CSDH is Burr-hole drainage with or without rinse Considering the poor outcomes of CSDH in aged patients, and ambiguous prediction of recurrence in many sides of recurrent CSDHs who have been analyzed, new effective therapies are needed for those CSDHs who are predicated to have poor prognosis for surgery and/or have a higher risk of recurrence. Statins, which is the first-line treatment for patients with high cholesterol and coronary heart disease. However, statins are still not solely limited in the treatment of these diseases. It has been demonstrated that statins could improve CSDH due to its effect of regulation of angiogenesis and inflammation. In this review, in order to provide potential new treatment for CSDH we summarize the recent findings of statins in CSDH in order to try to clarify the mechanisms of this effect.

Determining speaker attributes from stress-affected speech in emergency situations with hybrid SVM-DNN architecture

In the millions of emergency reporting calls made each year, about a quarter are non-emergencies. To avoid responding to such situations, forensic examination of the reported situation in the presence of speech as evidence has become an indispensable requirement for emergency response centers. Caller profile information like gender, age, emotional state, transcript, and contextual sounds determined from emergency calls, may be highly beneficial for their sophisticated forensic analysis. However, callers reporting emergency situations often express emotional stress which cause variations in speech production. Furthermore, low voice quality, and background noise make it very difficult to efficiently recognize caller attributes in such unconstrained environments. To overcome limitations of traditional classification systems in such situations, a hybrid two-stage classification scheme is proposed in this paper. Our framework consist of an ensemble of support vector machines (e-SVM) and deep neural networks (DNN) in a cascade. The first stage e-SVM consists of two models discriminatively trained on normal and stressful speech from emergency calls. Deep neural network forming the second stage of classification pipeline, is utilized only in case of ambiguous prediction results from the first stage. The adaptive nature of this two stage classification scheme helps achieve efficiency and high performance. Experiments conducted with a large dataset affirm the suitability of proposed architecture for efficient real-time speaker attribute recognition. The framework is evaluated for gender recognition from emergency calls in the presence of emotions and background noise. The framework yields significant performance improvements in comparison with other similar state-of-the-art gender recognition approaches.

Target definition for shipwreck hunting.

The research described in the present article was implemented to define the locations of two World War II shipwrecks, the German raider Kormoran, and the Australian light cruiser HMAS Sydney. The paper describes the long and complex trail that led through inefficient oceanographic prediction to ambiguous historical prediction involving a single report and on to precise cognitive prediction based on nine reports from more than 70 survivors, a process that yielded a single target position or “mean” just 2.7 NM (nautical miles) from the wreck of Kormoran. Prediction for the position of the wreck of Sydney opened with wishful thinking that she had somehow reached the coast more than 100 NM away when cognitive analysis of the survivor's reports actually provided the basis for accurate prediction in a position near to the wreck of Kormoran. In the account provided below, the focus on cognitive procedures emerged from, first, a review of a sample of the shipwreck hunts, and, second, growing awareness of the extraordinarily rich database available for this search, and the extent to which it was open to cognitive analysis. This review touches on both the trans-disciplinary and the cognitive or intra-disciplinary issues that so challenged the political entities responsible for supervising of the search for the wrecks of Kormoran and Sydney. One of the theoretical questions that emerged from these debate concerns the model of expertise advanced by Collins (2013). The decomposability of alleged forms of expertise is revealed as a fundamental problem for research projects that might or might not benefit from trans-disciplinary research. Where expertise can be decomposed for operational purposes, the traditional dividing lines between experts and novices, and fools for that matter, are much harder to discern, and require advanced and scientifically informed review.

Abstract 4364: Characterization of mucin-like 1 (MUCL1) in breast cancer and its novel role as a potent activator of cell proliferation

Abstract MUCL1 was first identified as a breast-specific gene in 2001. Based on its highly restricted mRNA expression in normal tissues, and its continued expression during breast tumorigenesis and progression (expressed in >90% of breast cancers), MUCL1 is an attractive tumor associated antigen as a potential therapeutic target. However, very little is known about the cellular location and biological functions of the MUCL1 protein, which will have a major impact on its druggability. Here we describe our efforts to fully characterize the cellular localization of MUCL1 and discover its functional roles in breast cancer. Protein sequence analysis yielded an ambiguous prediction as MUCL1 contains a peptide signal sequence for targeting to the ER/Golgi secretory pathway, but it could also act as a weak transmembrane domain. Flow cytometry and cell surface protein biotinylation studies were performed to assess whether it could be a surface membrane protein. The MUCL1 protein was found in both whole cell lysates and conditioned mediums of native cancer cell lines or ectopically expressing cell lines. However, cell surface expression was not detectable in either system. Interestingly, intracellular protein expression did not necessarily correspond with levels of secreted protein. Phenotypic assays revealed a strong dependence on MUCL1 for cell proliferation, but not cell migration/invasion, in several breast cancer cell lines. We further explored the mechanism by which MUCL1 regulated cell growth by carrying out cell cycle analysis and assessing expression of cell cycle markers. Knockdown of MUCL1 by siRNA triggered a G1/S phase growth arrest along with decreased Cyclin D1 expression and increased expression of the cyclin-dependent kinase inhibitors (CDKIs) p21 and p27. Finally, we investigated the impact of MUCL1 loss on kinase signaling pathways in breast cancer cells through phospho-kinase array profiling. Strong phospho-FAK, JNK and c-Jun signals were detected in control cell lines which were abrogated by MUCL1 siRNA and confirmed by Western blot. Our data indicate that MUCL1 is not a membrane bound protein, but is secreted by some breast cancer cells, while others only express high levels of intracellular MUCL1. We are the first to identify an important role for MUCL1 in the proliferation of breast cancer cells, likely mediated via the FAK/JNK signaling pathway. These data suggest a potential use for targeting this protein in breast cancer. Citation Format: Sarah J. Conley, Emily Bosco, David Tice, Robert Hollingsworth, Ronald Herbst, Zhan Xiao. Characterization of mucin-like 1 (MUCL1) in breast cancer and its novel role as a potent activator of cell proliferation. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4364. doi:10.1158/1538-7445.AM2015-4364

Predicting the Next Accessed Web Page Using Markov Model and PageRank

In this highly competitive world, only a user can decide the future of a web-based business. Success of any web business directly depends on services provided to the users and the level of satisfaction experienced by users. Prediction algorithm can be a dominating factor in web businesses that includes offers to customers on selling items, web site personalization and web recommendations etc. Understanding user's behaviour through Markov model can be a simple approach, but ambiguous prediction of next accessed web page is the main drawback of this approach. Therefore a method is proposed in this paper, in which ambiguous prediction problem can be resolved using Web PageRank along with Markov Model. Experimental result shows reduced number of ambiguous predictions after applying PageRank method.

Quantum corrections to the semiclassical Hartree-Fock theory of a harmonically trapped Bose gas

Using the phase-space expansion of the thermodynamical distribution functions we provide a general and systematic method for including effects beyond the local-density approximation to the semiclassical Hartree-Fock theories. We illustrate the method by applying it to the case of a strictly two-dimensional, harmonically trapped Bose gas. Thereby, we address the ambiguous prediction of the Hartree-Fock approximation, namely, whether a fixed number of trapped atoms undergoes Bose-Einstein condensation or not. We also investigate the dependence of the critical temperature on the interaction strength.

Taxes, Risk Aversion, and the Size of the Underground Economy

Non‐parametric regression analysis was used to investigate the relationship between the effective tax rate and the relative size of the underground economy, using New Zealand data. The underlying theoretical framework is established, and it suggests an ambiguous prediction regarding the sign of the relationship we are studying. However, our non‐parametric empirical analysis, which also allows for the non‐stationarity of the time‐series data, produces a positive and ‘S‐shaped’ relationship, and this supports earlier empirical studies that imposed such functional forms. The estimated model is used to simulate the effects of hypothetical tax changes on the size of the New Zealand underground economy, and to draw policy conclusions.

Vibration of Thermally Stressed Composite Plates with and Without Cutouts

predictedskin-frictioncoefe cients C f.Thesuperiorperformanceof theMCHmodel,instrongcontrasttothatoftheOCHmodel,isonce more ascertained. Apparently, the ambiguous prediction regarding theOCHmodelisattributabletoshortcomingsinthe y + dependence viscous damping functions employed. Conclusion The potential importance of the cross diffusion together with the viscous damping functions is conspicuous. The modie cation introduced with the Chien model is profoundly convenient because it circumvents the defect entangled with the model to a greater extent.TheMCHmodelaccountsforthenear-wallandlow-Reynoldsnumber effects emanating from the physical requirements.