Ani Dong

Bio: Ani Dong is an academic researcher from Dongguan University of Technology. The author has contributed to research in topics: Autoencoder & Anaerobic oxidation of methane. The author has an hindex of 2, co-authored 2 publications receiving 23 citations.

Efficient and High-quality Recommendations via Momentum-incorporated Parallel Stochastic Gradient Descent-Based Learning

Abstract: A recommender system (RS) relying on latent factor analysis usually adopts stochastic gradient descent (SGD) as its learning algorithm. However, owing to its serial mechanism, an SGD algorithm suffers from low efficiency and scalability when handling large-scale industrial problems. Aiming at addressing this issue, this study proposes a momentum-incorporated parallel stochastic gradient descent (MPSGD) algorithm, whose main idea is two-fold: a) implementing parallelization via a novel data-splitting strategy, and b) accelerating convergence rate by integrating momentum effects into its training process. With it, an MPSGD-based latent factor (MLF) model is achieved, which is capable of performing efficient and high-quality recommendations. Experimental results on four high-dimensional and sparse matrices generated by industrial RS indicate that owing to an MPSGD algorithm, an MLF model outperforms the existing state-of-the-art ones in both computational efficiency and scalability.

A Fast Deep AutoEncoder for High-Dimensional and Sparse Matrices in Recommender Systems

Abstract: A latent factor analysis (LFA)-based model has outstanding performance in extracting desired patterns from High-dimensional and Sparse (HiDS) data for building a recommender systems. However, they mostly fail in acquiring non-linear features from an HiDS matrix. An AutoEncoder (AE)-based model can address this issue efficiently, but it requires filling unknown data of an HiDS matrix with pre-assumptions that leads to the following limitations: a) prefilling unknown data of an HiDS matrix might skew its known data distribution to generate ridiculous recommendations; and b) incorporating a deep AE-style structure to improve its representative learning ability. Experimental results on three HiDS matrices from real recommender systems show that an FDAE-based model significantly outperforms state-of-the-art recommenders in terms of recommendation accuracy. Meanwhile, its computational efficiency is comparable with the most efficient recommenders with the help of parallelization.

Exploration of single Fe atom supported on anatase TiO2(001) for methane oxidation: A DFT study

Abstract: Partial oxidation of methane is a promising alternative strategy for methanol production under mild reaction conditions; however, significant challenges hinder the development of appropriate catalysts. In this study, based on first-principles calculations, we demonstrate that a single Fe atom supported on anatase TiO2(001) provides double active sites (Fe and Ti5C) to activate gas-phase O2 and form O-assisted intermediates. The triple state Fe-O/TiO2(001) system exhibited better activity for methane activation (ΔGmax = 1.02 eV). Our findings offer new insights into the design of non-noble-3d transition metal single-atom catalysts on TiO2(001) for partial methane oxidation via an inexpensive O2 oxidant under mild reaction conditions.

Highly-Accurate Community Detection via Pointwise Mutual Information-Incorporated Symmetric Non-Negative Matrix Factorization

Abstract: Community detection, aiming at determining correct affiliation of each node in a network, is a critical task of complex network analysis. Owing to its high efficiency, Symmetric and Non-negative Matrix Factorization (SNMF) is frequently adopted to handle this task. However, existing SNMF models mostly focus on a network's first-order topological information described by its adjacency matrix without considering the implicit associations among involved nodes. To address this issue, this study proposes a Pointwise mutual information-incorporated and Graph-regularized SNMF (PGS) model. It uses a) Pointwise Mutual Information to quantify implicit associations among nodes, thereby completing the missing but crucial information among critical nodes in a uniform way; b) graph-regularization to achieve precise representation of local topology, and c) SNMF to implement efficient community detection. Empirical studies on eight real-world social networks generated by industrial applications demonstrate that a PGS model achieves significantly higher accuracy gain in community detection than state-of-the-art community detectors.

Symmetric Nonnegative Matrix Factorization-Based Community Detection Models and Their Convergence Analysis

Abstract: Community detection is a popular yet thorny issue in social network analysis. A symmetric and nonnegative matrix factorization (SNMF) model based on a nonnegative multiplicative update (NMU) scheme is frequently adopted to address it. Current research mainly focuses on integrating additional information into it without considering the effects of a learning scheme. This study aims to implement highly accurate community detectors via the connections between an SNMF-based community detector's detection accuracy and an NMU scheme's scaling factor. The main idea is to adjust such scaling factor via a linear or nonlinear strategy, thereby innovatively implementing several scaling-factor-adjusted NMU schemes. They are applied to SNMF and graph-regularized SNMF models to achieve four novel SNMF-based community detectors. Theoretical studies indicate that with the proposed schemes and proper hyperparameter settings, each model can: 1) keep its loss function nonincreasing during its training process and 2) converge to a stationary point. Empirical studies on eight social networks show that they achieve significant accuracy gain in community detection over the state-of-the-art community detectors.

A Novel Approach to Large-Scale Dynamically Weighted Directed Network Representation

Abstract: A dynamically weighted directed network (DWDN) is frequently encountered in various big data-related applications like a terminal interaction pattern analysis system (TIPAS) concerned in this study. It consists of large-scale dynamic interactions among numerous nodes. As the involved nodes increase drastically, it becomes impossible to observe their full interactions at each time slot, making a resultant DWDN High Dimensional and Incomplete (HDI). An HDI DWDN, in spite of its incompleteness, contains rich knowledge regarding involved nodes various behavior patterns. To extract such knowledge from an HDI DWDN, this paper proposes a novel Alternating direction method of multipliers (ADMM)-based Nonnegative Latent-factorization of Tensors (ANLT) model. It adopts three-fold ideas: a) building a data density-oriented augmented Lagrangian function for efficiently handling an HDI tensors incompleteness and nonnegativity; b) splitting the optimization task in each iteration into an elaborately designed subtask series where each one is solved based on the previously solved ones following the ADMM principle to achieve fast convergence; and c) theoretically proving that its convergence is guaranteed with its efficient learning scheme. Experimental results on six DWDNs from real applications demonstrate that the proposed ANLT outperforms state-of-the-art models significantly in both computational efficiency and prediction accuracy.

CARM: Confidence-aware recommender model via review representation learning and historical rating behavior in the online platforms

Abstract: The recommendation systems in the online platforms often suffer from the rating data sparseness and information overload issues. Previous studies on this topic often leverage review information to construct an accurate user/item latent factor. To address this issue, we propose a novel confidence-aware recommender model via review representation learning and historical rating behavior in this article. It is motived that ratings are consistent with reviews in terms of user preferences, and reviews often contain misleading comments (e.g., fake good reviews, fake bad reviews). To this end, the interaction latent factor of user and item in the framework is constructed by exploiting review information interactivity. Then, the confidence matrix, which measures the relationship between the rating outliers and misleading reviews, is employed to further improve the model accuracy and reduce the impact of misleading reviews on the model. Furthermore, the loss function is constructed by maximum a posteriori estimation theory. Finally, the mini-batch gradient descent algorithm is introduced to optimize the loss function. Experiments conducted on four real-world datasets empirically demonstrate that our proposed method outperforms the state-of-the-art methods. The proposed method also further promotes the application in learning resource adaptation. The source Python code will be available upon request.

Fast and Accurate Non-Negative Latent Factor Analysis of High-Dimensional and Sparse Matrices in Recommender Systems

Abstract: A fast non-negative latent factor (FNLF) model for a high-dimensional and sparse (HiDS) matrix adopts a Single Latent Factor-dependent, Non-negative, Multiplicative and Momentum-incorporated Update (SLF-NM ² U) algorithm, which enables its fast convergence. It is crucial to achieve a rigorously theoretical proof regarding its fast convergence, which has not been provided in prior research. Aiming at addressing this critical issue, this work theoretically proves that with an appropriately chosen momentum coefficient, SLF-NM ² U enables the fast convergence of an FNLF model in both continuous and discrete time cases. Empirical analysis of HiDS matrices generated by representative industrial applications provides empirical evidences for the theoretical proof. Hence, this study represents an important milestone in the field of HiDS matrix analysis.