Search

Agriculture is essential for global income, poverty reduction, and food security, with crop yield being a crucial measure in this field. Traditional crop yield prediction methods, reliant on subjective assessments such as farmers' experiences, tend to be error-prone and lack precision across vast farming areas, especially in data-scarce regions. Recent advancements in data collection, notably through high-resolution sensors and the use of deep learning (DL), have significantly increased the accuracy and breadth of agricultural data, providing better support for policymakers and administrators. In our study, we conduct a systematic literature review to explore the application of DL in crop yield forecasting, underscoring its growing significance in enhancing yield predictions. Our approach enabled us to identify 92 relevant studies across four major scientific databases: the Directory of Open Access Journals (DOAJ), the Institute of Electrical and Electronics Engineers (IEEE), the Multidisciplinary Digital Publishing Institute (MDPI), and ScienceDirect. These studies, all empirical research published in the last eight years, met stringent selection criteria, including empirical validity, methodological clarity, and a minimum quality score, ensuring their rigorous research standards and relevance. Our in-depth analysis of these papers aimed to synthesize insights on the crops studied, DL models utilized, key input data types, and the specific challenges and prerequisites for accurate DL-based yield forecasting. Our findings reveal that convolutional neural networks and Long Short-Term Memory are the dominant deep learning architectures in crop yield prediction, with a focus on cereals like wheat (Triticum aestivum) and corn (Zea mays). Many studies leverage satellite imagery, but there is a growing trend towards using Unmanned Aerial Vehicles (UAVs) for data collection. Our review synthesizes global research, suggests future directions, and highlights key studies, acknowledging that results may vary across different databases and emphasizing the need for continual updates due to the evolving nature of the field. [ABSTRACT FROM AUTHOR]

Burn‐in is an effective method to screen out early failures of electronic devices. Typically, this is achieved by operating the devices under accelerated stress conditions. This paper focuses on a burn‐in concept where a random sample of devices is drawn out of the running production, put to burn‐in, and investigated for early failures. This procedure is called burn‐in study. In parallel, as long as the burn‐in study is ongoing, all other produced devices are subjected to burn‐in screening. In this article, new flexible sampling plans for burn‐in studies are introduced. These are based on the progress of these studies and defined quality targets. Furthermore, these sampling plans enable fast burn‐in time reductions and various time reduction strategies. From a statistical point of view, this requires to combine the proportion of early failures in a population with their lifetime distribution function. The new model is illustrated by case studies and simulations. It contributes to burn‐in cost reductions, while controlling quality levels at the same time. [ABSTRACT FROM AUTHOR]

Semiconductor devices must fulfill highest quality standards since they are used in safety relevant applications. Bad devices, namely devices which are not fully functional after the production are scrapped immediately and not delivered to the customer. Unfortunately, among the remaining ones there are still devices with increased risk, prone to infant mortality. To minimize the chance of delivering such risk devices, statistical screening methods are applied to Front-End data to detect suspicious devices, represented as statistical outliers. Dependent on the technology, different measurements are suitable for screening. Nevertheless, it is assumed that the measurements contain not only signal noise but also a hidden signal identifying risk devices. Therefore, this paper proposes to use the Independent Component Analysis as a data transformation to separate informative and non-informative content from the electrical measurements. [ABSTRACT FROM AUTHOR]

The space-time interpolation of precipitation has significant contribution to river control, reservoir operations, forestry interest and flash flood watches etc. The changes in environmental covariates and spatial covariates make space-time estimation of precipitation a challenging task. In the present paper, we use a generalized additive model with Gaussian link function to account for the effect of covariates; the resulting output is partitioned into two parts; trend component and residual component. The trend component is modeled on the basis of spatial artificial neural network (SANN) architecture. The residual component is assumed to be a spatio-temporal random field and is modeled using Le and Zidek (2006) hierarchical Bayesian interpolation (HBI) method. The separable stationary space-time nested covariance model and purely spatial non-stationary non-parametric covariance model for interpolation of the residual component are used. For the interpolation of the amount of precipitation at ungauged locations the interpolated residual components for ungauged locations are added to the respective interpolated trend components. The results of two covariance functions are compared by means of cross-validations and suggest that HBI including covariates provides minimum mean square prediction error if the nested spatio-temporal stationary covariance model is used. [ABSTRACT FROM AUTHOR]