We analyze a compression scheme for large data sets that randomly keeps small percentage of the components each sample. The benefit is output sparse matrix, and therefore, subsequent processing, such as principal component analysis (PCA) or K-means, significantly faster, especially in distributed-data setting. Furthermore, sampling single-pass applicable to streaming data. mechanism variant previous methods proposed literature combined with randomized preconditioning smooth provide...

A fundamental task in machine learning involves visualizing high-dimensional data sets that arise high-impact application domains. When considering the context of large imbalanced data, this problem becomes much more challenging. In paper, t-Distributed Stochastic Neighbor Embedding (t-SNE) algorithm is used to reduce dimensions an earthquake engineering related set for visualization purposes. Since greatly affect accuracy classifiers, we employ Synthetic Minority Oversampling Technique...

Machine learning techniques were used to predict tensile properties of material extrusion-based additively manufactured parts made with Technomelt PA 6910, a hot melt adhesive. An adaptive data generation technique, specifically an active process based on the Gaussian regression algorithm, was employed enable prediction limited training data. After three rounds collection, machine models linear regression, ridge and K-nearest neighbors tasked predicting for test dataset, which consisted...

Generating low-rank approximations of kernel matrices that arise in nonlinear machine learning techniques holds the potential to significantly alleviate memory and computational burdens. A compelling approach centers on finding a concise set exemplars or landmarks reduce number similarity measure evaluations from quadratic linear concerning data size. However, key challenge is regulate tradeoffs between quality resource consumption. Despite volume research this area, current understanding...

Dictionary learning algorithms design a dictionary that is specifically tailored to enable sparse representation of given set training signals. In turn, the increased sparsity signals with respect this enables significantly improved performance in variety state-of-the-art signal processing tasks, e.g. compressive sensing. However, while these typically assume all data fully available, may not be case practice. fact, high cost acquiring each or sheer amount acquired motivate us take sensing...

A pivotal aspect in the design of neural networks lies selecting activation functions, crucial for introducing nonlinear structures that capture intricate input-output patterns. While effectiveness adaptive or trainable functions has been studied domains with ample data, like image classification problems, significant gaps persist understanding their influence on accuracy and predictive uncertainty settings characterized by limited data availability. This research aims to address these...

This work evaluates the stability of Boostan earth dam by investigating its long-term performance and interpreting measured data. To measure response, several sensitive locations are instrumented. process includes measuring various quantities such as pore water pressure, level, internal stress ratios using inspection devices ordinary Casagrande piezometers, total pressure cells. The recorded data shows that is in good agreement with initial (stable) design condition. installed piezometers...

This paper proposes a systematic approach for the seismic design of 2D concrete dams. As opposed to traditional method which does not optimize dam cross-section, proposed engine offers optimal one based on predefined constraints. A large database about 24,000 simulations is generated transient simulation dam-foundation-water system. The includes over 150 various shapes, water levels, and material properties, as well 160 different ground motion records. Automated machine learning (AutoML)...

The Nystrom method is a popular technique for generating low-rank approximations of kernel matrices that arise in many machine learning problems. approximation quality the depends crucially on number selected landmark points and selection procedure. In this paper, we introduce randomized algorithm scalable to large high-dimensional data sets. proposed performs K-means clustering low-dimensional random projections set thus leads significant savings Our theoretical results characterize...

Compressive sensing allows us to recover signals that are linearly sparse in some basis from a smaller number of measurements than traditionally required. However, it has been shown many classes images or video can be more efficiently modeled as lying on nonlinear manifold, and hence described non-linear function few underlying parameters. Recently, there growing interest using these manifold models reduce the required compressive measurements. complexity an obstacle their use efficient data...

This paper focuses on the estimation of sample covariance matrix from low-dimensional random projections data known as compressive measurements. In particular, we present an unbiased estimator to extract structure measurements obtained by a general class projection matrices consisting i.i.d. zero-mean entries and finite first four moments. contrast previous works, make no structural assumptions about underlying such being low-rank. fact, our analysis is based non-Bayesian setting which...

Uncertainty quantification in complex engineering problems is challenging because of necessitating large numbers expensive model evaluations. This paper proposes a two-stage framework for developing accurate machine learning-based surrogate models structural engineering. The studied numerical considers aleatory and epistemic uncertainties, i.e., ground motion features material properties. Our framework's first step trains classification algorithms on the collected data from our with...