Deep autoencoder-based fuzzy c-means for topic detection

Murfi, Hendri; Rosaline, Natasha; Hariadi, Nora

doi:10.1016/j.array.2021.100124

Cited by 12 publications

(5 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The complexity and overlap facing many multi-dimensional data made traditional clustering algorithms useless, so it was necessary to develop clustering techniques that could deal with the overlap between clusters and form homogeneous groups according to fuzzy logic. The first algorithm to address this interference was Fuzzy C Means (FCM) 22 , which aims to collect large data in the form of new, more homogeneous groups based on determining degrees of belonging to the targeted cases. This algorithm was developed by Dunn, Bezdek In 1974, by developing the partition matrix in the K-Means clustering algorithm while determining the degree of fuzzing, then the objective function was obtained, which represents the minimization sum of square errors 𝒙 𝒊𝒋 : value of observation i at dimension j. 𝒗 𝒋𝒌 : represents the center of cluster k of dimension j. 𝝋(𝒙 𝒊𝒋 ; 𝒗 𝒌 ): is a measure of similarity or difference of the observation value𝒙 𝒊𝒋 from its cluster center 𝒗 𝒋𝒌 .…”

Section: Fuzzy C-means Clusteringmentioning

confidence: 99%

Enhancing Fuzzy C-Means Clustering with a Novel Standard Deviation Weighted Distance Measure

Mohammed,

Ashour

2024

Baghdad Sci.J

View full text Add to dashboard Cite

The aim of this paper is to present a new approach to address the Fuzzy C Mean algorithm, which is considered one of the most important and famous algorithms that addressed the phenomenon of uncertainty in forming clusters according to the overlap ratios. One of the most important problems facing this algorithm is its reliance primarily on the Euclidean distance measure, and by nature, the situation is that this measure makes the formed clusters take a spherical shape, which is unable to contain complex or overlapping cases. Therefore, this paper attempts to propose a new measure of distance, where we were able to derive a formula for the variance of the fuzzy cluster to be entered as a weight on the Euclidean Distance (WED) formula. Moreover, the calculation was processed partitions matrix through the use of the K-Means algorithm and creating a hybrid environment between the fuzzy algorithm and the sharp algorithm. To verify what was presented, experimental simulation was used and then applied to reality using environmental data for the physical and chemical examination of water testing stations in Basra Governorate. It was proven through the experimental results that the proposed distance measure Weighted Euclidean distance had the advantage over improving the work of the HFCM algorithm through the criterion (Obj_Fun, Iteration, Min_optimization, good fit clustering and overlap) when (c = 2,3) and according to the simulation results, c = 2 was chosen to form groups for the real data, which contributed to determine the best objective function (23.93, 22.44, 18.83) at degrees of fuzzing (1.2, 2, 2.8), while according to the degree of fuzzing (m = 3.6), the objective function for Euclidean Distance (ED) was the lowest, but the criteria were (Iter. = 2, Min_optimization = 0 and ) which confirms that (WED) is the best.

show abstract

Section: Fuzzy C-means Clusteringmentioning

confidence: 99%

Enhancing Fuzzy C-Means Clustering with a Novel Standard Deviation Weighted Distance Measure

Mohammed,

Ashour

2024

Baghdad Sci.J

View full text Add to dashboard Cite

show abstract

“…This type of data representation is called clusteringfriendly representation [18]. Another method used for deep text clustering is the autoencoder neural network [19] [20], which is one of the most prominent unsupervised representation learning algorithms [21]. This neural network creates a non-linear mapping from the data space to a latent space in order to reduce the dimensions.…”

Section: Research Backgroundmentioning

confidence: 99%

Neural Network Meaningful Learning Theory and its Application for Deep Text Clustering

Zafarani-Moattar,

Kangavari,

Rahmani

2024

IEEE Access

View full text Add to dashboard Cite

In this paper, a new theory to train neural networks is presented which is called "Neural Network Meaningful Learning" (NNMeL) theory. According to this theory, meaningful learning in an artificial neural network occurs when concepts are learned separately but get related to each other. This theory is theoretically supported by "Azobel's cognitive theory of learning". According to Azobel's theory, the most important effective factor in learning is previous learning, and meaningful learning occurs when a person consciously relates new knowledge to what they already knew. Also, a new model named "Deep Clustering based on Sentence Similarity" (DCSS) is proposed for topic detection. This model proposes to use similarity of sentences to produce sentence representation instead of using autoencoder and training it based on denoising. Also, a trainable framework based on NNMeL is presented. Many experiments conducted for evaluation. First of all, an experiment is arranged to check the correctness of the NNMeL theory. The results confirm that training ANN according to this theory (training concepts one after another), gets better results than training all concepts together. In the following experiments, the DCSS model, which is based on sentence similarity, is compared with another autoencoder based method not only by evaluation metrics but also by human evaluators. The results indicate a 6.6% improvement in accuracy and better human-evaluators satisfaction. Finally, the proposed model is compared with ten other methods in topic detection application. The results show that the proposed model is superior to the rest of the methods.

show abstract

“…problems. Autoencoders (AE) [1] and GAN [2] are two techniques for making deepfakes. Applications like FakeApp, DeepFaceLab, DFaker, and DeepFaketf have all adopted autoencoder technology.…”

Section: A Fake-image Generation and Deepfake Datasetsmentioning

confidence: 99%

“…For example, Reddit user "deepfakes" employs deep learning algorithms to distribute female face photos into pornographic videos. There is a significant uproar in response to this instance of defamation and disrespect [1]. Moreover, BuzzFeed has ever produced a video with former US President Barack Obama that he gives a speech about the problems of deepfakes on individuals and society as a whole [2].…”

Section: Introductionmentioning

confidence: 99%

Classification of Deepfake Images Using a Novel Explanatory Hybrid Model

Kerenalli,

Yendapalli,

Chinnaiah

2023

CommIT (Communication and Information Technology) Journal

View full text Add to dashboard Cite

In court, criminal investigations and identity management tools, like check-in and payment logins, face videos, and photos, are used as evidence more frequently. Although deeply falsified information may be found using deep learning classifiers, block-box decisionmaking makes forensic investigation in criminal trials more challenging. Therefore, the research suggests a three-step classification technique to classify the deceptive deepfake image content. The research examines the visual assessments of an EfficientNet and Shifted Window Transformer (SWinT) hybrid model based on Convolutional Neural Network (CNN) and Transformer architectures. The classifier generality is improved in the first stage using a different augmentation. Then, the hybrid model is developed in the second step by combining the EfficientNet and Shifted Window Transformer architectures. Next, the GradCAM approach for assessing human understanding demonstrates deepfake visual interpretation. In 14,204 images for the validation set, there are 7,096 fake photos and 7,108 real images. In contrast to focusing only on a few discrete face parts, the research shows that the entire deepfake image should be investigated. On a custom dataset of real, Generative Adversarial Networks (GAN)-generated, and human-altered web photos, the proposed method achieves an accuracy of 98.45%, a recall of 99.12%, and a loss of 0.11125. The proposed method successfully distinguishes between real and manipulated images. Moreover, the presented approach can assist investigators in clarifying the composition of the artificially produced material.

show abstract

Deep autoencoder-based fuzzy c-means for topic detection

Cited by 12 publications

References 28 publications

Enhancing Fuzzy C-Means Clustering with a Novel Standard Deviation Weighted Distance Measure

Enhancing Fuzzy C-Means Clustering with a Novel Standard Deviation Weighted Distance Measure

Neural Network Meaningful Learning Theory and its Application for Deep Text Clustering

Classification of Deepfake Images Using a Novel Explanatory Hybrid Model

Contact Info

Product

Resources

About