Error-free Data Set Research Articles

High-risk AI systems in the EU Commission’s Proposal for an AI Act

On April 21, 2021, the EU Commission published a comprehensive proposal for an AI Act, which is intended to be the world s first set of regulations to create the legal framework for the development, use and sale of AI systems. The focus of the proposal is on the regulation of so-called high-risk AI systems. The article elaborates the obligations of providers, importers, traders and users of high-risk AI systems. These are part of IT and product safety law and - even if not explicitly addressed in the proposal - have an impact on contract law and tort law, which the article also illuminates. It is consequent that the Commission holds both providers and users responsible for subjecting high-risk AI systems to monitoring and security obligations throughout their entire life cycle. The requirements and obligations of the Proposed AI Act can ensure an increased level of legal certainty with regard to the use of AI systems in civil law. Nevertheless, the Proposed AI Act also reveals considerable weaknesses. First of all, the fuzzy definition of AI systems, which also includes software that does not pose any AI-specific risks, should be emphasized. Furthermore, the security requirements of the Proposed AI Act are partly too vague or difficult to implement in practice, which is particularly clear from the requirement for complete and error-free data sets and the retailer s inspection obligations. The relief for small and medium-sized companies is still too vague. In its current form, the Proposed AI Act could therefore ensure that Europe as an AI location is extremely cautious in developing and using AI. With a view to the wide range of applications of AI, it can be assumed that the AI Regulation will have just as much of an impact on the European and global market as the GDPR.

Impact of Stereo Camera Calibration to Object Accuracy in Multimedia Photogrammetry

Camera calibration via bundle adjustment is a well-established standard procedure in single-medium photogrammetry. When using standard software and applying the collinearity equations in multimedia photogrammetry, the effects of refractive interfaces are compensated in an implicit form, hence by the usual parameters of interior orientation. This contribution analyses different calibration strategies for planar bundle-invariant interfaces. To evaluate the effects of implicitly modelling the refractive effects within bundle adjustment, synthetic error-free datasets are simulated. The behaviour of interior, exterior, and relative orientation parameters is analysed using synthetic datasets free of underwater imaging effects. A shift of the camera positions of 0.2% of the acquisition distance along the optical axis can be observed. The relative orientation of a stereo camera shows systematic effects when the angle of convergence varies. The stereo baseline increases by 1% at 25° convergence. Furthermore, the interface is set up at different distances to the camera. When the interface is at 50% distance assuming a parallel camera setup, the stereo baseline also increases by 1%. It becomes clear that in most cases the implicit modelling is not suitable for multimedia photogrammetry due to geometrical errors (scaling) and absolute positioning errors. Explicit modelling of the refractive interfaces is implemented into a bundle adjustment and is also used to analyse calibration parameters and deviations in object space. Real experiments show that it is difficult to separate the effects of implicit modelling, since other effects, such as poor image measurements, affect the final result. However, trends can be seen, and deviations are quantified.

Influence of uncertain identification of triggering rainfall on the assessment of landslide early warning thresholds

Abstract. Uncertainty in rainfall datasets and landslide inventories is known to have negative impacts on the assessment of landslide-triggering thresholds. In this paper, we perform a quantitative analysis of the impacts of uncertain knowledge of landslide initiation instants on the assessment of rainfall intensity–duration landslide early warning thresholds. The analysis is based on a synthetic database of rainfall and landslide information, generated by coupling a stochastic rainfall generator and a physically based hydrological and slope stability model, and is therefore error-free in terms of knowledge of triggering instants. This dataset is then perturbed according to hypothetical reporting scenarios that allow simulation of possible errors in landslide-triggering instants as retrieved from historical archives. The impact of these errors is analysed jointly using different criteria to single out rainfall events from a continuous series and two typical temporal aggregations of rainfall (hourly and daily). The analysis shows that the impacts of the above uncertainty sources can be significant, especially when errors exceed 1 day or the actual instants follow the erroneous ones. Errors generally lead to underestimated thresholds, i.e. lower than those that would be obtained from an error-free dataset. Potentially, the amount of the underestimation can be enough to induce an excessive number of false positives, hence limiting possible landslide mitigation benefits. Moreover, the uncertain knowledge of triggering rainfall limits the possibility to set up links between thresholds and physio-geographical factors.

Calibration of the Fundamental Diagram Based on Loop and Probe Data

The fundamental diagram is a key component of traffic flow. This diagram describes equilibrium traffic states and their propagation on a traffic network. Knowledge of traffic parameters is of paramount importance to understand traffic properties and characteristics. It is also critical to calibrate the elements of dynamic traffic flow simulation models and reproduce traffic states on road networks. This paper describes the development of a method for estimating fundamental diagram parameters that combines loop data and probe data. Loop data are considered boundary conditions of the problem. The travel times between any points located in the loop can be estimated on the basis of the fundamentals of the kinematic wave theory. The optimal fundamental diagram parameters are computed so that the discrepancy between estimated travel times and the actual travel times determined with probe vehicles is minimal. The method is validated by the use of simulated error-free data. The results demonstrate the accuracy of the method when it is applied to an error-free data set. The method is then implemented with realistic data, that is, data that were aggregated and noised beforehand. The robustness of the method is demonstrated, and the results are encouraging for the development of an algorithm that calibrates fundamental diagram parameters online and automatically.

Harnessing Computational Biology for Exact Linear B-Cell Epitope Prediction: A Novel Amino Acid Composition-Based Feature Descriptor.

Proteins embody epitopes that serve as their antigenic determinants. Epitopes occupy a central place in integrative biology, not to mention as targets for novel vaccine, pharmaceutical, and systems diagnostics development. The presence of T-cell and B-cell epitopes has been extensively studied due to their potential in synthetic vaccine design. However, reliable prediction of linear B-cell epitope remains a formidable challenge. Earlier studies have reported discrepancy in amino acid composition between the epitopes and non-epitopes. Hence, this study proposed and developed a novel amino acid composition-based feature descriptor, Dipeptide Deviation from Expected Mean (DDE), to distinguish the linear B-cell epitopes from non-epitopes effectively. In this study, for the first time, only exact linear B-cell epitopes and non-epitopes have been utilized for developing the prediction method, unlike the use of epitope-containing regions in earlier reports. To evaluate the performance of the DDE feature vector, models have been developed with two widely used machine-learning techniques Support Vector Machine and AdaBoost-Random Forest. Five-fold cross-validation performance of the proposed method with error-free dataset and dataset from other studies achieved an overall accuracy between nearly 61% and 73%, with balance between sensitivity and specificity metrics. Performance of the DDE feature vector was better (with accuracy difference of about 2% to 12%), in comparison to other amino acid-derived features on different datasets. This study reflects the efficiency of the DDE feature vector in enhancing the linear B-cell epitope prediction performance, compared to other feature representations. The proposed method is made as a stand-alone tool available freely for researchers, particularly for those interested in vaccine design and novel molecular target development for systems therapeutics and diagnostics: https://github.com/brsaran/LBEEP.

SCLAP: An Adaptive Boosting Method for Predicting Subchloroplast Localization of Plant Proteins

Chloroplasts are organelles found in plant system and other photosynthetic eukaryotes. Since a large number of essential pathways take place in this organelle, proteins in the chloroplast are considered vital. Therefore, knowledge about the subchloroplast localization of the chloroplast proteins will provide precise information in understanding its interaction within the chloroplast. To address this, an AdaBoost-based prediction system to predict the subchloroplast localization of chloroplast proteins (SCLAP) was developed. It integrates three different sequence-based features for prediction, beside the addition of similarity-based module for significant improvement in prediction performance. SCLAP achieved an overall accuracy of 89.3% in jackknife cross-validation test against the benchmark dataset, which was considered highest among existing tools and equals the SubIdent, and 85.9% accuracy in new error-free dataset. Evaluation of SCLAP with the independent dataset, five-fold cross-validation, and their corresponding receiver operator characteristic curve analysis demonstrated the SCLAP's efficient performance. SCLAP is the webserver implementation of our algorithm written in PERL. The server can be used to predict the subchloroplast localization of chloroplast proteins ( http://sclap.bicpu.edu.in/predict.php ).

Different approaches for quantifying ventilation distribution and lung tissue properties by functional EIT

We investigated five different methods which can be applied to quantitatively construct functional tomograms of the lungs. The focus was on the sensitivity of functional tomograms to errors in acquired data. To quantify this sensitivity, theoretical, error-free data sets of well-known properties were artificially generated based on a ‘living thorax model’. Physiological time courses and a typical distribution of errors caused by a typical Goe-MF II EIT system were used for the calculations which encompassed a range up to 50 times greater than the initial error level (4 µVrms max–400 µVrms max). Additionally, low-pass filtering and principal component analysis (PCA) were used to quantify the effect of preprocessing the raw data. The results demonstrate that all methods based on fitting the local to the global time course were superior to the common functional tomograms utilizing standard deviation or maximum and minimum detection. Ventilation distribution was best quantified by the so-called VT methods. Filling capacity—a lung tissue property—was least dependent on increasing error levels. The errors introduced by filtering are significant with respect to a quantitative analysis of ventilation distribution. A preprocessing of raw data by applying a PCA performed well on the data sets which had been constructed but were, nonetheless, realistic. This approach appears to be highly promising for application on real data which is known to be erroneous.

Development and implementation of high-throughput SNP genotyping in barley

BackgroundHigh density genetic maps of plants have, nearly without exception, made use of marker datasets containing missing or questionable genotype calls derived from a variety of genic and non-genic or anonymous markers, and been presented as a single linear order of genetic loci for each linkage group. The consequences of missing or erroneous data include falsely separated markers, expansion of cM distances and incorrect marker order. These imperfections are amplified in consensus maps and problematic when fine resolution is critical including comparative genome analyses and map-based cloning. Here we provide a new paradigm, a high-density consensus genetic map of barley based only on complete and error-free datasets and genic markers, represented accurately by graphs and approximately by a best-fit linear order, and supported by a readily available SNP genotyping resource.ResultsApproximately 22,000 SNPs were identified from barley ESTs and sequenced amplicons; 4,596 of them were tested for performance in three pilot phase Illumina GoldenGate assays. Data from three barley doubled haploid mapping populations supported the production of an initial consensus map. Over 200 germplasm selections, principally European and US breeding material, were used to estimate minor allele frequency (MAF) for each SNP. We selected 3,072 of these tested SNPs based on technical performance, map location, MAF and biological interest to fill two 1536-SNP "production" assays (BOPA1 and BOPA2), which were made available to the barley genetics community. Data were added using BOPA1 from a fourth mapping population to yield a consensus map containing 2,943 SNP loci in 975 marker bins covering a genetic distance of 1099 cM.ConclusionThe unprecedented density of genic markers and marker bins enabled a high resolution comparison of the genomes of barley and rice. Low recombination in pericentric regions is evident from bins containing many more than the average number of markers, meaning that a large number of genes are recombinationally locked into the genetic centromeric regions of several barley chromosomes. Examination of US breeding germplasm illustrated the usefulness of BOPA1 and BOPA2 in that they provide excellent marker density and sensitivity for detection of minor alleles in this genetically narrow material.

Eliminating the effect of overlapping crowns from aerial inventory estimates

When a forest stand is inventoried from aerial photographs or with the use of laser scanners, only those trees that are not covered by a crown of a taller tree can be observed. This means that the probability of a tree being observable from the air depends on its relative height among the trees in the stand and on the crown areas of the taller trees. This study derives a mathematical formula for the probability of a tree being observable from the air according to certain assumptions about the censoring process in a forest stand where trees are randomly located without spatial autocorrelation in tree size. After presenting this formula, different approaches are proposed for the correction of the censoring effect upon the observed distribution of crown areas. The methods could change the aim of individual-tree recognition from finding all trees of the stand to finding a set of trees for which the stated assumptions about the censoring process are valid as far as possible. The methods performed well in a demonstration with simulated, error-free data sets, for which all stated assumptions were valid.

Quantitative Analysis of Variability and Uncertainty with Known Measurement Error: Methodology and Case Study

The appearance of measurement error in exposure and risk factor data potentially affects any inferences regarding variability and uncertainty because the distribution representing the observed data set deviates from the distribution that represents an error-free data set. A methodology for improving the characterization of variability and uncertainty with known measurement errors in data is demonstrated in this article based on an observed data set, known measurement error, and a measurement-error model. A practical method for constructing an error-free data set is presented and a numerical method based upon bootstrap pairs, incorporating two-dimensional Monte Carlo simulation, is introduced to address uncertainty arising from measurement error in selected statistics. When measurement error is a large source of uncertainty, substantial differences between the distribution representing variability of the observed data set and the distribution representing variability of the error-free data set will occur. Furthermore, the shape and range of the probability bands for uncertainty differ between the observed and error-free data set. Failure to separately characterize contributions from random sampling error and measurement error will lead to bias in the variability and uncertainty estimates. However, a key finding is that total uncertainty in mean can be properly quantified even if measurement and random sampling errors cannot be separated. An empirical case study is used to illustrate the application of the methodology.

Approaches to Cleaning Data Sets

Since each of the methods described has its costs and benefits, more than one method should be used. The combination of methods helps the researcher to obtain an error-free data set. Ideally, value-to-value verification for all data sets is preferred but the multiple entry method is more efficient for large data sets. Each researcher has a definition of large and small data sets. One definition of a large data set is one containing more than 250 records. Such a definition will not be appropriate for all investigators. Therefore, each researcher should balance the allocation of resources needed for a method against the confidence required for the data.