Abstract:A Bayesian Deep Restricted Boltzmann-Kohonen architecture for data clustering termed as DRBM-ClustNet is proposed. This core-clustering engine consists of a Deep Restricted Boltzmann Machine (DRBM) for processing unlabeled data by creating new features that are uncorrelated and have large variance with each other. Next, the number of clusters are predicted using the Bayesian Information Criterion (BIC), followed by a Kohonen Network-based clustering layer. The processing of unlabeled data is done in three stages for efficient clustering of the non-linearly separable datasets. In the first stage, DRBM performs non-linear feature extraction by capturing the highly complex data representation by projecting the feature vectors of $d$ dimensions into $n$ dimensions. Most clustering algorithms require the number of clusters to be decided a priori, hence here to automate the number of clusters in the second stage we use BIC. In the third stage, the number of clusters derived from BIC forms the input for the Kohonen network, which performs clustering of the feature-extracted data obtained from the DRBM. This method overcomes the general disadvantages of clustering algorithms like the prior specification of the number of clusters, convergence to local optima and poor clustering accuracy on non-linear datasets. In this research we use two synthetic datasets, fifteen benchmark datasets from the UCI Machine Learning repository, and four image datasets to analyze the DRBM-ClustNet. The proposed framework is evaluated based on clustering accuracy and ranked against other state-of-the-art clustering methods. The obtained results demonstrate that the DRBM-ClustNet outperforms state-of-the-art clustering algorithms.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is some major drawbacks existing in the existing clustering algorithms in data clustering, such as the need to pre - specify the number of clusters, being prone to converge to local optimal solutions, and having poor clustering accuracy on non - linear data sets. To solve these problems, the author proposes a data clustering architecture based on Bayesian Deep Restricted Boltzmann - Kohonen Network (DRBM - ClustNet). ### Specific Problems and Solutions 1. **Pre - specifying the number of clusters**: - **Problem**: Most clustering algorithms need to determine the number of clusters in advance, which is often difficult to achieve in practical applications. - **Solution**: In the paper, the Bayesian Information Criterion (BIC) is used to automatically predict the number of clusters, thus avoiding the need to manually specify the number of clusters. 2. **Converging to local optimal solutions**: - **Problem**: Traditional clustering algorithms (such as K - means) are prone to fall into local optimal solutions, resulting in unsatisfactory clustering results. - **Solution**: By using the Deep Restricted Boltzmann Machine (DRBM) for feature extraction, new, uncorrelated and features with large variances are generated, thereby improving the global optimization ability of the clustering algorithm. 3. **Poor clustering accuracy on non - linear data sets**: - **Problem**: Traditional clustering algorithms perform poorly when dealing with non - linear data sets, especially on high - dimensional data. - **Solution**: DRBM - ClustNet efficiently clusters non - linear data sets through multi - stage processing. First, DRBM performs non - linear feature extraction; second, BIC is used to predict the number of clusters; finally, the Kohonen Network is used for clustering. ### Model Architecture 1. **Feature Extraction Stage**: - DRBM is used to process unlabeled data to generate new features. DRBM projects the feature vector from d - dimension to n - dimension by capturing highly complex data representations. 2. **Clustering Number Prediction Stage**: - BIC is used to automatically predict the number of clusters. BIC evaluates the fitness of different models by statistical methods and selects the best number of clusters. 3. **Clustering Stage**: - The Kohonen Network is used to cluster the data after feature extraction. The Kohonen Network is a self - organizing mapping network and can effectively handle high - dimensional data. ### Experimental Verification The paper uses two synthetic data sets, 15 benchmark data sets in the UCI Machine Learning Repository and four image data sets to verify the effectiveness of DRBM - ClustNet. The experimental results show that DRBM - ClustNet is superior to other existing clustering algorithms in terms of clustering accuracy and robustness, especially when dealing with non - linear data sets. ### Summary DRBM - ClustNet effectively solves the deficiencies of existing clustering algorithms by combining the Deep Restricted Boltzmann Machine and the Kohonen Network, providing an automated, efficient and accurate data clustering method. This method has significant advantages when dealing with complex, non - linear data sets.

DRBM-ClustNet: A Deep Restricted Boltzmann-Kohonen Architecture for Data Clustering

Deep Clustering and Visualization for End-to-End High-Dimensional Data Analysis.

Evolving Restricted Boltzmann Machine-Kohonen Network for Online Clustering

Deep learning-based clustering approaches for bioinformatics

Learning deep representation and discriminative features for clustering of multi-layer networks

Deep Reinforcement Clustering

Deep Density-based Image Clustering

Unsupervised hierarchical clustering using the learning dynamics of RBMs

DeepDPM: Deep Clustering With an Unknown Number of Clusters

DRSCDM: A Novel Density-Related Clustering for Complex High-Dimensional Data Streams

Learning a Bi-Directional Discriminative Representation for Deep Clustering

Detection of Cloned Attacks in Connecting Media using Bernoulli RBM_RF Classifier (BRRC)

An RBF Neural Network Clustering Algorithm Based on K-Nearest Neighbor

Rectified factor networks for biclustering of omics data

A Dual Adaptive Assignment Approach for Robust Graph-Based Clustering

BDEC:Brain Deep Embedded Clustering model

Latent Subspace Clustering based on Deep Neural Networks

Deep Learning with Nonparametric Clustering

DeepCluE: Enhanced Image Clustering via Multi-layer Ensembles in Deep Neural Networks

Deep Discriminative Clustering Analysis

Robust Deep K-Means: an Effective and Simple Method for Data Clustering.