Unsupervised random forest for affinity estimation

This paper presents an unsupervised clustering random-forest-based metric for affinity estimation in large and high-dimensional data. The criterion used for node splitting during forest construction can handle rank-deficiency when measuring cluster compactness. The binary forest-based metric is extended to continuous metrics by exploiting both the common traversal path and the smallest shared parent node.The proposed forest-based metric efficiently estimates affinity by passing down data pairs in the forest using a limited number of decision trees. A pseudo-leaf-splitting (PLS) algorithm is introduced to account for spatial relationships, which regularizes affinity measures and overcomes inconsistent leaf assign-ments. The random-forest-based metric with PLS facilitates the establishment of consistent and point-wise correspondences. The proposed method has been applied to automatic phrase recognition using color and depth videos and point-wise correspondence. Extensive experiments demonstrate the effectiveness of the proposed method in affinity estimation in a comparison with the state-of-the-art.

A FINITE ELEMENT MODEL FOR COMPRESSIVE ICE LOADS BASED ON A MOHR-COULOMB MATERIAL AND THE NODE SPLITTING TECHNIQUE

This paper presents a finite element model for the simulation of ice-structure interaction problems, which are dominated by crushing. The failure mode of ice depends significantly on the strain rate. At low strain rates the ice behaves ductile, whereas at high strain rates ice reacts in brittle mode. This paper focuses on the brittle mode, which is the dominating mode for ship-ice interactions. A multitude of numerical approaches for the simulation of ice can be found in the literature. Nevertheless, the literature approaches do not seem suitable for the simulation of continuous ice-structure interaction processes at low and high confinement ratios in brittle mode. Therefore, this paper seeks to simulate the ice-structure interaction with the finite element method (FEM). The objective of the here introduced Mohr-Coulomb Nodal Split (MCNS) model is to represent the essential material behavior of ice in an efficient formulation. To preserve mass and energy as much as possible, the node splitting technique is applied, instead of the frequently used element erosion technique. The intention of the presented model is not to reproduce individual cracks with high accuracy, because this is not possible with a reasonable element size, due to the large number of crack fronts forming during the ice-structure interaction process. To validate the findings of the model, the simulated maximum ice forces and contact pressures are compared with ice-extrusion and double pendulum tests. During validation, the MCNS model shows a very good agreement with these experimental values.

A Finite Element Model for Compressive Ice Loads Based on a Mohr-Coulomb Material and the Node Splitting Technique

This paper presents a finite element model for the simulation of ice-structure interaction problems, which are dominated by crushing at low and medium confinement ratios. The failure mode of ice depends significantly on the strain rate. At very low impact velocities the ice behaves ductile, whereas at high velocities the ice reacts in brittle mode. This paper focuses on the brittle mode, which is the dominating mode for ship-ice interactions. A multitude of numerical approaches for the simulation of ice can be found in the literature. Nevertheless, the literature approaches do not seem suitable for the simulation of continuous ice-structure interaction processes at low and medium confinement ratios in brittle mode. Therefore, this paper seeks to simulate the ice-structure interaction with the FE method. To preserve mass and energy as much as possible, the node splitting technique is applied, instead of the often used element erosion technique. The intention of the presented model is not to reproduce individual cracks with high accuracy, because this is not possible with a reasonable element size, due to the large number of crack fronts forming during the ice-structure interaction process. The objective of the here introduced Mohr-Coulomb Nodal Split (MCNS) model is to represent the essential material behavior of ice in a efficient formulation. To validate the findings of the model, the simulated maximum ice forces and contact pressures are compared with experiments.

REPRESENTATION VARIETIES OF ALGEBRAS WITH NODES

We study the behaviour of representation varieties of quivers with relations under the operation of node splitting. We show how splitting a node gives a correspondence between certain closed subvarieties of representation varieties for different algebras, which preserves properties like normality or having rational singularities. Furthermore, we describe how the defining equations of such closed subvarieties change under the correspondence. By working in the ‘relative setting’ (splitting one node at a time), we demonstrate that there are many nonhereditary algebras whose irreducible components of representation varieties are all normal with rational singularities. We also obtain explicit generators of the prime defining ideals of these irreducible components. This class contains all radical square zero algebras, but also many others, as illustrated by examples throughout the paper. We also show that this is true when irreducible components are replaced by orbit closures, for a more restrictive class of algebras. Lastly, we provide applications to decompositions of moduli spaces of semistable representations of certain algebras.

BoostTree and BoostForest for Ensemble Learning

Bootstrap aggregating (Bagging) and boosting are two popular ensemble learning approaches, which combine multiple base learners to generate a composite model for more accurate and more reliable performance. They have been widely used in biology, engineering, healthcare, etc. This article proposes BoostForest, which is an ensemble learning approach using BoostTree as base learners and can be used for both classification and regression. BoostTree constructs a tree model by gradient boosting. It achieves high randomness (diversity) by sampling its parameters randomly from a parameter pool, and selecting a subset of features randomly at node splitting. BoostForest further increases the randomness by bootstrapping the training data in constructing different BoostTrees. BoostForest outperformed four classical ensemble learning approaches (Random Forest, Extra-Trees, XGBoost and LightGBM) on 34 classification and regression datasets. Remarkably, BoostForest has only one hyper-parameter (the number of BoostTrees), which can be easily specified. Our code is publicly available, and the proposed ensemble learning framework can also be used to combine many other base learners.

The research of destruction of functionally-graded materials subjected to impact loading

A numerical analysis on impact response of multilayer plates and plates with a gradient substrate against steel projectile perforation was made. The shear strength was varied in the substrate within a certain range. The behavior of bodies is modeled by an elastic-plastic, porous, compressible medium, taking into account shock-wave phenomena and fragmentary fracture of materials. A numerical lagrangian method with modified node splitting algorithms was used. Good agreement between the computed and experimental results was obtained. During perforation, pattern of destruction of all plates has been investigated. The results show impact resistance of plates with a gradient substrate was greater than the homogeneous steel one, but less than multilayer ones. However, the impact resistance of multilayer plates is explained by the pinching effect of the layers.

Experimental tests and numerical simulations of ballistic impact on laminated glass

In this study, we investigate double-laminated glass plates under ballistic impact through experimental tests and numerical simulations. The experimental tests are used to determine the ballistic limit velocity and curve for the laminated glass targets, and to create a basis for comparison with numerical simulations. We tested two different glass pane configurations: (1) one double-laminated glass plate, and (2) two layers of double-laminated glass plates separated by an airgap. In the numerical study, we used finite element simulations that employed higher order elements and 3D node splitting to predict the residual velocities of the bullets in the experiments. Node splitting enabled modelling of fracture by element separation and was employed for the glass parts. The material and fracture models that we used for the glass and the PVB parts were simplified, but the numerical predictions proved to be in excellent agreement with the experimental results.

Classification Algorithm in Data Mining Based on Maximum Exponential Class Counts Technique

A new split attribute measure for decision tree node split during decision tree creation is proposed. The new split measure consists of the sum of class counts of distinct values of categorical attributes in the dataset. Larger counts induce larger partitions and smaller trees there by favors to the determination of the best spit attribute. The new split attribute measure is termed as maximum exponential class counts (MECC). Experiment results obtained over several UCI machine learning categorical datasets predominantly indicate that the decision tree models created based on the proposed MECC node split attribute technique provides better classification accuracy results and smaller trees in size than the decision trees created using popular gain ratio, normalized gain ratio and gini-index measures. The experimental results are mainly focused on performing and analyzing the results from the node splitting measures alone.

Study on Cargo Products’ Layout Optimization of Truck Railway Line

Cargo products’ layout is one of the problems that needs an important decision in freight railroads. The purpose of optimization of the cargo products’ layout on railway lines is to make a rational cargo products’ plan to satisfy customers’ diversified transport demands. In this paper, a service network is designed using the method of node splitting. Then, an optimized model is built for the products’ layout problem on busy main railway lines. The immune clone-variable neighborhood algorithm is used to solve the model. Finally, a numerical example is given to verify the model and algorithm. The result shows that the model and algorithm are of good practical applicability.