A Consensus Approach to Improve NMF Document Clustering

Febrissy, Mickaël; Nadif, Mohamed

doi:10.1007/978-3-030-44584-3_14

Cited by 3 publications

(3 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the optimizing strategy of ( 6) still faces the challenges of high complexity. Specifically, once the matrix Z is learned, calculating the Laplacian matrix requires O(n 2 ) computational complexity during optimizing V. Compared with standard NMF with O(n) complexity, this will limit the scalability of model (6). Importantly, the graph Laplacian regularizer is a Euclidean-based distance metric, which can be reformulated as…”

Section: Motivation and Objective Functionmentioning

confidence: 99%

See 1 more Smart Citation

Globality constrained adaptive graph regularized non‐negative matrix factorization for data representation

Sun

Wang

Guo

et al. 2022

IET Image Processing

View full text Add to dashboard Cite

Benefiting from the good physical interpretations and low computational complexity, nonnegative matrix factorization (NMF) has attracted wide attentions in data representation learning tasks. Some graph-based NMF approaches make the learned representation encode the topological structure by the local graph Laplacian regularizer, which improves the discriminant ability of data representation. However, the performance of graph-based NMF methods depend heavily on the quality of the predefined graph and the complexity of models is high. Here, a globality constrained adaptive graph regularized non-negative matrix factorization for data representation (GCAG-NMF) model is proposed, which not only uses the self-representation characteristics of data to learn an adaptive graph to describe the sample relationship more accurately, but also proposes a graph factorization technique to reduce the complexity of the model and improve the discriminative ability of data representation. Then, an iterative optimizing strategy with low complexity and strict convergence guarantee is developed to optimize the objective function. Experimental results on some databases demonstrate the effectiveness of the proposed model.

show abstract

Section: Motivation and Objective Functionmentioning

confidence: 99%

“…This non-negativity constraints result in the parts-based representation of NMF. In practice, NMF has been applied to many applications, such as document clustering [6], image segmentation [7], text mining [8], image clustering [9,10] and biological data mining [11].…”

Section: Introductionmentioning

confidence: 99%

Globality constrained adaptive graph regularized non‐negative matrix factorization for data representation

Sun

Wang

Guo

et al. 2022

IET Image Processing

View full text Add to dashboard Cite

show abstract

“…IR has presented several techniques for the processing and retrieval of relevant documents [18]. As in the last few years, we have observed document clustering as a proliferation solution in applying IR based on the assumption that if a document is pertinent to a query, other documents of that cluster can pertain to that query too [19]. It is believed that data storage indexing can be improved via grouping XML documents together which will have a positive effect on the retrieval process [20].…”

Section: Introductionmentioning

confidence: 99%

XML Clustering Framework Based on Document Content and Structure in a Heterogeneous Digital Library

Samadi,

(Corresponding Author)

2023

MJCS

View full text Add to dashboard Cite

As textually published information is increasing in digital libraries, efficient retrieval methods are required. Textual documents in a digital library are available in various structures and contents. It is possible to represent these documents with hierarchical levels of granularity when these are organized in XML structure to improve precision by focused retrieval. By this means, contextual elements of each document can be retrieved from a known structure. One solution for retrieving these elements is clustering from a combination of Content and Structural similarities. To achieve this, a novel two-level clustering framework based on Content and Structure is proposed. The framework decomposes a document into meaningful structural units and analyzes all its rich text in its own structure. The quality of the proposed framework was experimented on a heterogeneous XML document collection, having varieties of data sources, structures, and content, be represented as a sample of a real digital library. This collection was made with capabilities to test all of our objectives. The clustering results were evaluated by the Entropy criterion. Finally, the Content and Structure clustering was compared with the usual clustering based on the Content Only to prove the efficacy of considering structural features against the existing Content Only methods in the retrieval process. The total Entropy results of the two-level Content and Structural clustering are almost twice better than the Content Only clustering approach. Consequently, the proposed framework has the ability to improve Information Retrieval systems from two points of view: i) considering the structural aspect of text-rich documents in the retrieval process, and ii) replacing the document-level retrieval with the element-level retrieval.

show abstract

Nonnegative matrix factorization with Wasserstein metric-based regularization for enhanced text embedding

Li,

Wang,

et al. 2024

PLoS ONE

View full text Add to dashboard Cite

Text embedding plays a crucial role in natural language processing (NLP). Among various approaches, nonnegative matrix factorization (NMF) is an effective method for this purpose. However, the standard NMF approach, fundamentally based on the bag-of-words model, fails to utilize the contextual information of documents and may result in a significant loss of semantics. To address this issue, we propose a new NMF scheme incorporating a regularization term based on the Wasserstein metric, which quantifies an approximation of a word-context matrix to leverage semantic information. Since the word-context matrix is typically symmetric and positive definite (SPD), the Wasserstein metric used in the regularization is related to a natural SPD manifold structure. We then build upon this manifold structure to integrate an explicit expression of the Wasserstein metric into the NMF framework. Additionally, we introduce modifications to the gradient computation to adapt three fundamental classes of numerical algorithms from ordinary NMF to tackle the Wasserstein-regularized NMF (NMF-WR) problem. To demonstrate the effectiveness of the NMF-WR model and the proposed algorithms, we conduct experiments on popular datasets, focusing on topic modeling and document clustering. The results indicate that the NMF-WR model exhibits superior performance compared to other conventional NMF models in processing text embeddings. These findings suggest that this novel NMF-WR framework not only enhances semantic representation but also underscores a commitment to the model’s reliability and interpretability.

show abstract

A Consensus Approach to Improve NMF Document Clustering

Cited by 3 publications

References 26 publications

Globality constrained adaptive graph regularized non‐negative matrix factorization for data representation

Globality constrained adaptive graph regularized non‐negative matrix factorization for data representation

XML Clustering Framework Based on Document Content and Structure in a Heterogeneous Digital Library

Nonnegative matrix factorization with Wasserstein metric-based regularization for enhanced text embedding

Contact Info

Product

Resources

About