Engineering Papers⌕ Search

Engineering topics

Song, Huan

Publications and source records attributed to Song, Huan.

Systems and methods for customizing kernel machines with deep neural networks

A method including receiving an input data set. The input data set can include one of a feature domain set or a kernel matrix. The method also can include constructing dense embeddings using: (i) Nyström approximations on the input data set when the input data set comprises the kernel matrix, and (ii) clustered Nyström approximations on the input data set when the input data set comprises the feature domain set. The method additionally can include performing representation learning on each of the dense embeddings using a multi-layer fully-connected network for each of the dense embeddings to generate latent representations corresponding to each of the dense embeddings. The method further can include applying a fusion layer to the latent representations corresponding to the dense embeddings to generate a combined representation. The method additionally can include performing classification on the combined representation. Other embodiments of related systems and methods are also disclosed.

Song, Huan↗

GrAMME: Semisupervised Learning Using Multilayered Graph Attention Models

Modern data analysis pipelines are becoming increasingly complex due to the presence of multiview information sources. While graphs are effective in modeling complex relationships, in many scenarios, a single graph is rarely sufficient to succinctly represent all interactions, and hence, multilayered graphs have become popular. Though this leads to richer representations, extending solutions from the single-graph case is not straightforward. Consequently, there is a strong need for novel solutions to solve classical problems, such as node classification, in the multilayered case. In this article, we consider the problem of semisupervised learning with multilayered graphs. Though deep network embeddings, e.g., DeepWalk, are widely adopted for community discovery, we argue that feature learning with random node attributes, using graph neural networks, can be more effective. To this end, we propose to use attention models for effective feature learning and develop two novel architectures, GrAMME-SG and GrAMME-Fusion, that exploit the interlayer dependences for building multilayered graph embeddings. Using empirical studies on several benchmark data sets, we evaluate the proposed approaches and demonstrate significant performance improvements in comparison with the state-of-the-art network embedding strategies. The results also show that using simple random features is an effective choice, even in cases where explicit node attributes are not available.

97 MATHEMATICS AND COMPUTING↗

Effective Removal of Anionic Re(VII) by Surface-Modified Ti 2 CT x MXene Nanocomposites: Implications for Tc(VII) Sequestration

Environmental contamination by 99 Tc(VII) from radioactive wastewater streams is of particular concern due to the long half-life of 99 Tc and high mobility of pertechnetate. Here in this paper, we report a novel MXene-polyelectrolyte nanocomposite with three-dimensional networks for enhanced removal of perrhenate, which is pertechnetate simulant. The introduction of poly(diallyldimethylammonium chloride) (PDDA) regulates the surface charge and improves the stability of Ti 2 CT x nanosheet, resulting in Re(VII) removal capacity of up to 363 mg g –1 , and fast sorption kinetics. The Ti 2 CT x /PDDA nanocomposite furthermore exhibits good selectivity for ReO 4 – when competing anions (such as Cl – and SO 4 2– ) coexist at a concentration of 1800 times. The immobilization mechanism was confirmed as a sorption-reduction process by batch sorption experiments and X-ray photoelectron spectroscopy. The pH-dependent reducing activity of Ti 2 CT x /PDDA nanocomposite toward Re(VII) was clarified by X-ray absorption spectroscopy. As the pH increases, the local environment gradually changes from octahedral-coordinated Re(IV) to tetrahedral-coordinated Re(VII). The overall results suggest that Ti 2 CT x /PDDA nanocomposite may be a promising candidate for efficient elimination of Tc contamination. The reported surface modification strategy might result in applications of MXene-based materials in environmental remediation of other oxidized anion pollutants.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗

Efficient U(VI) Reduction and Sequestration by Ti 2 CT x MXene

Although reduction of highly mobile U(VI) to less soluble U(IV) has been long considered an effective approach to in situ environmental remediation of uranium, candidate reducing agents are largely limited to Fe-based materials and microbials. The importance of titanium-containing compounds in natural uranium ore deposits suggests a role for titanium in uranium migration. In this report, for the first time, a two-dimensional transition metal carbide, Ti 2 CT x , is shown to efficiently remove uranium via a sorption-reduction strategy. Batch experiments demonstrate that Ti 2 CT x exhibits excellent U(VI) removal over a wide pH range, with an uptake capacity of 470 mg g –1 at pH 3.0. The mechanism for U(VI) to U(IV) reduction by Ti 2 CT x was deciphered by X-ray absorption spectroscopy and diffraction and photoelectron spectroscopy. The reduced U(IV) species at low pH is identified as mononuclear with bidendate binding to the MXene substrate. At near-neutral pH, nanoparticles of the UO 2+x phase adsorb to the substrate with some Ti 2 CT x transformed to amorphous TiO 2 . A subsequent in-depth study suggests Ti 2 CT x materials may be potential candidates for permeable reactive barriers in the treatment of wastewaters from uranium mining. Finally, this work highlights reduction-induced immobilization of U(VI) by Ti 2 CT x MXene including a pH-dependent reduction mechanism that might promote applications of titanium-based materials in the elimination of other oxidized contaminants.

36 MATERIALS SCIENCE↗

Optimizing Kernel Machines Using Deep Learning

Building highly non-linear and non-parametric models is central to several state-of-the-art machine learning systems. Kernel methods form an important class of techniques that induce a reproducing kernel Hilbert space (RKHS) for inferring non-linear models through the construction of similarity functions from data. These methods are particularly preferred in cases where the training data sizes are limited and when prior knowledge of the data similarities is available. Despite their usefulness, they are limited by the computational complexity and their inability to support end-to-end learning with a task specific objective. On the other hand, deep neural networks have become the de facto solution for end-to-end inference in several learning paradigms. In this article, we explore the idea of using deep architectures to perform kernel machine optimization, for both computational efficiency and end-to-end inferencing. To this end, we develop the DKMO (Deep Kernel Machine Optimization) framework, that creates an ensemble of dense embeddings using Nystrõm kernel approximations and utilizes deep learning to generate task-specific representations through the fusion of the embeddings. Intuitively, the filters of the network are trained to fuse information from an ensemble of linear subspaces in the RKHS. Furthermore, we introduce the kernel dropout regularization to enable improved training convergence. Finally, we extend this framework to the multiple kernel case, by coupling a global fusion layer with pre-trained deep kernel machines for each of the constituent kernels. Using case studies with limited training data, and lack of explicit feature sources, we demonstrate the effectiveness of our framework over conventional model inferencing techniques.

97 MATHEMATICS AND COMPUTING↗