Clustering is an unsupervised learning task that groups data points by their inherent similarities. Nonautomatic clustering algorithms face significant challenges when the true number of clusters is unknown or changes dynamically, as they require this number to be predefined. This paper provides a comprehensive review of automatic clustering algorithms specifically designed to handle such uncertainty. In this paper, these algorithms are systematically classified based on three key perspectives: clustering framework (classical vs. deep), clustering strategy (e.g., density-based, model based, graph-theoretic, subspace methods), and the use of labeled data (unsupervised vs. semi-supervised). We analyze each algorithm based on its core principles, key contributions, strengths, and limitations. Furthermore, we address the current challenges in this area and propose future research directions to enhance the scalability, robustness, and effectiveness of automatic clustering algorithms.

Variational graph auto-encoders (VGAEs) are a key tool for node clustering, but existing models face several significant challenges. These challenges include a mismatch between inference and generative models after incorporating the clustering inductive bias, as well as posterior collapse (PC), where latent representations become overly influenced by the prior distribution. In addition, in existing VGAEs, noisy clustering assignments lead to the feature randomness (FR) challenge, while the strong tradeoff between clustering accuracy and reconstruction quality results in the feature drift (FD) problem. To address these issues, we propose a multiscale contrastive VGAE (MCVGAE). This multiscale model combines cluster-level and graph-level contrastive learning with proximity-level and cluster-level self-supervised methods. MCVGAE improves the alignment between the hidden space and the data distribution and prevents PC. Moreover, it reduces FR and FD more effectively than existing techniques. Achieving impressive accuracy scores of 79.09% on Cora, 90.04% on ACM, 75.12% on Pubmed, 72.7% on Citeseer, 74.11% on DBLP, and 59.79% on Wiki clearly demonstrates the superiority of MCVGAE over 30 state-of-the-art methods.

Nowadays, sentiment analysis is concerned with identifying and analysing text sentiment. Sentiment analysis has been used in many fields because of its applications in various domains. In the last decade, with the success of machine learning and deep learning methods, many machine- and deep-based sentiment classification have been developed and performed well on various issues. Moreover, word embeddings are important for machine learning and deep learning models since they provide input features in downstream language tasks. This paper presents a comprehensive review of word embeddings and deep learning models. Additionally, we conduct an experimental study of sentiment classification using various deep learning models and word embeddings, in which five deep learning models with four embedding techniques are compared on eight benchmark datasets. In other words, 20 models are evaluated on datasets. Finally, we discuss the performance of models from different perspectives.

Anomaly detection (AD) has witnessed substantial advancements in recent years due to the increasing need for identifying outliers in various engineering applications that undergo environmental adaptations. Consequently, researchers have focused on developing robust AD methods to enhance system performance. The primary challenge faced by AD algorithms lies in effectively detecting unlabeled abnormalities. This study introduces an adaptive evolutionary autoencoder (AEVAE) approach for AD in time-series data. The proposed methodology leverages the integration of unsupervised machine learning techniques with evolutionary intelligence to classify unlabeled data. The unsupervised learning model employed in this approach is the AE network. A systematic programming framework has been devised to transform AEVAE into a practical and applicable model. The primary objective of AEVAE is to detect and predict outliers in time-series data from unlabeled data sources. The effectiveness, speed, and functionality enhancements of the proposed method are demonstrated through its implementation. Furthermore, a comprehensive statistical analysis based on performance metrics is conducted to validate the advantages of AEVAE in terms of unsupervised AD.

Distributed global snapshot (DGS) is one of the fundamental protocols in distributed systems. It is used for different applications like collecting information from a distributed system and taking checkpoints for process rollback. The Chandy–Lamport protocol (CLP) is famous and well‐known for taking DGS. The main aim of this protocol was to generate consistent cuts without interrupting the regular operation of the distributed system. CLP was the origin of many future protocols and inspired them. The first aim of this paper is to propose a novel formal hierarchical parametric colored Petri net model of CLP. The number of constituting processes of the model is parametric. The second aim is to automatically generate a novel message sequence chart (MSC) to show detailed steps for each simulation run of the snapshot protocol. The third aim is model checking of the proposed formal model to verify the correctness of CLP and our proposed colored Petri net model. Having vital tools helps greatly to test the correct operation of the newly proposed distributed snapshot protocol. The proposed model of CLP can easily be used for visually testing the correct operation of the new future under‐development DGS protocol. It also permits formal verification of the correct operation of the new proposed protocol. This model can be used as a simple, powerful, and visual tool for the step‐by‐step run of the CLP, model checking, and teaching it to postgraduate students. The same approach applies to similar complicated distributed protocols.

Hand-written text recognition (HTR) is often carried out line-by-line: the decoding of text lines is carried out independently. This approach is known to deteriorate recognition accuracy of words and characters close to the line boundaries. The present study investigates this issue from the point of view of the language modeling component of the HTR system. Obviously, lack of linguistic context may be one of the reasons for loss of accuracy, but it certainly is not the only factor in play. We seek to clarify to which extent the problem can be influenced by the language modeling component of the system. We first discuss how to develop adapted language models which significantly improve HTR performance in general. We then focus on the deployment of methods to improve accuracy at line boundaries. The final result is an efficient approach which significantly improves HTR accuracy without changing the basic HTR system setup.

We present an intelligent sample selection approach to language model adaptation for handwritten text recognition, which exploits a combination of in-domain and out-of-domain data for construction of language models. In comparison to approaches proposed in the literature, our approach is characterized by a careful consideration of the criteria used for ranking samples and an innovative approach to sample selection which iteratively extends the training set for two language models. We propose two methods, in which agreement or disagreement of two ranking criteria (one for each language model) guides the selection of samples to add to the training sets of the models. Both approaches are shown to clearly outperform a strong baseline consisting of a carefully tuned interpolation of in-domain and out-of-domain language models.

In this paper, we consider the multiclass semi-supervised classification problem. A boosting algorithm is proposed to solve the multiclass problem directly. The proposed multiclass approach uses a new multiclass loss function, which includes two terms. The first term is the cost of the multiclass margin and the second term is a regularization term on unlabeled data. The regularization term is used to minimize the inconsistency between the pair wise similarity and the classifier predictions. It assigns the soft labels weighted with the similarity between unlabeled and labeled examples. We then derive a boosting algorithm, named CD-MSSBoost, from the proposed loss function using coordinate gradient descent. The derived algorithm is further used for learning optimal similarity function for a given data. Our experiments on a number of UCI datasets show that CD-MSSBoost outperforms the state-of-the-art methods to multiclass semi-supervised learning.

We present an algorithm for multiclass Semi-Supervised learning which is learning from a limited amount of labeled data and plenty of unlabeled data. Existing semi-supervised algorithms use approaches such as one-versus-all to convert the multiclass problem to several binary classification problems which is not optimal. We propose a multiclass semi-supervised boosting algorithm that solves multiclass classification problems directly. The algorithm is based on a novel multiclass loss function consisting of the margin cost on labeled data and two regularization terms on labeled and unlabeled data. Experimental results on a number of UCI datasets show that the proposed algorithm performs better than the state-of-the-art boosting algorithms for multiclass semi-supervised learning.

In this paper we present a new Multiclass semi-supervised learning algorithm that uses a base classifier in combination with a similarity function applied to all data to find a classifier that maximizes the margin and consistency over all data. A novel multiclass loss function is presented and used to derive the algorithm. We apply the algorithm to animal behavior recognition from accelerometer data. Animal-borne accelerometer data are collected from free-ranging animals and then labeled by a human expert. The resulting data are used to train a classifier. However, labeling is not easy from accelerometer data only and it is often not feasible to observe animals fitted with an accelerometer. All current approaches to this behavior recognition task use supervised or unsupervised learning. Since unlabeled data are easy to acquire and collect, a semi-supervised approach seems appropriate and reduces the human efforts for labeling. Experiments with accelerometer data collected from free-ranging gulls and benchmark UCI datasets show that the algorithm is effective and compares favorably with existing algorithms for multiclass semi-supervised learning.

Recently, Semi-Supervised learning algorithms such as co-training are used in many domains. In co-training, two classifiers based on different subsets of the features or on different learning algorithms are trained in parallel and unlabeled data that are classified differently by the classifiers but for which one classifier has large confidence are labeled and used as training data for the other. In this paper, a new form of co-training, called Ensemble-Co-Training, is proposed that uses an ensemble of different learning algorithms. Based on a theorem by Angluin and Laird that relates noise in the data to the error of hypotheses learned from these data, we propose a criterion for finding a subset of high-confidence predictions and error rate for a classifier in each iteration of the training process. Experiments show that the new method in almost all domains gives better results than the state-of-the-art methods.