Lactic Acid Bacteria From Saltwater Fish Samples: Probiotic Potential Characterization and Clustering Using Unsupervised Algorithms

This research aimed to isolate lactic acid bacteria (LAB) from the bowl of saltwater fish to assess their probiotic properties. Nineteen isolates of LAB including Lactobacillus plantarum, Lactobacillus acidophilus, Lactobacillus fermentum, Lactococcus lactis, Enterococcus hirae, Enterococcus lactis, Pediococcus pentosaceus, Pediococcus acidilactici, and Pediococcus lolli were recognized using molecular tools. All the isolates survived in the simulated conditions of the GI tract. Auto-aggregation ranged from 01.3 ± 0.5% to 82.6 ± 1.4% and hydrophobicity with toluene ranged from 3.7 ± 1.6% to 69.4 ± 1.3%, while the range of hydrophobicity with xylene was from 02.2 ± 1.6% to 56.4 ± 2.1%. All the isolates of lactobacilli, pediococci, enterococci, and lactococci indicated variable sensitivity and resistance towards clinical antibiotics. Non-neutralized cell free supernatant of isolates F12 and F15 showed antimicrobial activity against all the 8 evaluated enteric pathogens. Cluster analysis of identified potential probiotic bacteria based on heat-map and PCA methods also highlighted the priority of isolates F3, F7, F12, and F15 as bio-control agents in fishery industry. The findings of this study may essentially contribute to the understanding of the probiotic potential of LAB in saltwater fish, in order to access their probiotic characterization for use as biocontrol in fishery.

Fusion-Based Semantic Segmentation Using Deep Learning Architecture in Case of Very Small Training Dataset

Semantic segmentation is a pre-processing step in computer vision-based applications. It is the task of assigning a predefined class label to every pixel of an image. Several supervised and unsupervised algorithms are available to classify pixels of an image into predefined object classes. The algorithms, such as random forest and SVM are used to obtain the semantic segmentation. Recently, convolutional neural network (CNN)-based architectures have become popular for the tasks of object detection, object recognition, and segmentation. These deep architectures perform semantic segmentation with far better accuracy than the algorithms that were used earlier. CNN-based deep learning architectures require a large dataset for training. In real life, some of the applications may not have sufficient good quality samples for training of deep learning architectures e.g. medical applications. Such a requirement initiated a need to have a technique of effective training of deep learning architecture in case of a very small dataset. Class imbalance is another challenge in the process of training deep learning architecture. Due to class imbalance, the classifier overclassifies classes with large samples. In this paper, the challenge of training a deep learning architecture with a small dataset and class imbalance is addressed by novel fusion-based semantic segmentation technique which improves segmentation of minor and major classes.

Recommendation Systems on E-Learning and Social Learning: A Systematic Review

E-learning is renowned as one of the highly effective modalities of learning. Social learning, in turn, is considered to be of major importance as it promotes collaboration between learners. For properly managing learning resources, recommender systems have been implemented in e-learning to enhance learners' experience. Whilst recommender systems are of widespread concern in online learning, it is still unclear to educators how recommender systems can improve the learning process and have a positive impact on learning. This paper seeks to provide an overview of the recommender systems proposed in e-learning between 2007 and the first part of 2021. Out of 100 initially identified publications for the period between 2007 and the first part of 2021, 51 articles were included for final synthesis, according to specific criteria. The descriptive results show that most of the disciplines involved in educational recommender systems papers have approached e-learning in a general way without putting as much emphasis on social learning, and that recommender systems based on explicit feedbacks and ratings were the most frequently used in empirical studies. The synthesis of results presents several recommender systems types in e-learning: (1) Content-based recommender systems, (2) Collaborative-filtering recommender systems, (3) Hybrid recommender systems and (4) Recommender systems based on supervised and unsupervised algorithms. The conclusions reflect on the almost lack of critical reflection on the importance of addressing recommender systems in social learning and social educational networks in particular, especially as social learning has particular requirements, the weak databases size used in some research work, the importance of acknowledging the strengths and weaknesses of each type of recommender system in an educational context and the need for further exploration of implicit feedbacks more than explicit learners’ feedbacks for more accurate recommendations.

Unsupervised integration of single-cell multi-omics datasets with disparities in cell-type representation

Integrated analysis of multi-omics data allows the study of how different molecular views in the genome interact to regulate cellular processes; however, with a few exceptions, applying multiple sequencing assays on the same single cell is not possible. While recent unsupervised algorithms align single-cell multi-omic datasets, these methods have been primarily benchmarked on co-assay experiments rather than the more common single-cell experiments taken from separately sampled cell populations. Therefore, most existing methods perform subpar alignments on such datasets. Here, we improve our previous work Single Cell alignment using Optimal Transport (SCOT) by using unbalanced optimal transport to handle disproportionate cell-type representation and differing sample sizes across single-cell measurements. We show that our proposed method, SCOTv2, consistently yields quality alignments on five real-world single-cell datasets with varying cell-type proportions and is computationally tractable. Additionally, we extend SCOTv2 to integrate multiple ($M\geq2$) single-cell measurements and present a self-tuning heuristic process to select hyperparameters in the absence of any orthogonal correspondence information.

Meta-Learning to Improve Unsupervised Intrusion Detection in Cyber-Physical Systems

Artificial Intelligence (AI)- based classifiers rely on Machine Learning (ML) algorithms to provide functionalities that system architects are often willing to integrate into critical Cyber-Physical Systems (CPSs) . However, such algorithms may misclassify observations, with potential detrimental effects on the system itself or on the health of people and of the environment. In addition, CPSs may be subject to threats that were not previously known, motivating the need for building Intrusion Detectors (IDs) that can effectively deal with zero-day attacks. Different studies were directed to compare misclassifications of various algorithms to identify the most suitable one for a given system. Unfortunately, even the most suitable algorithm may still show an unsatisfactory number of misclassifications when system requirements are strict. A possible solution may rely on the adoption of meta-learners, which build ensembles of base-learners to reduce misclassifications and that are widely used for supervised learning. Meta-learners have the potential to reduce misclassifications with respect to non-meta learners: however, misleading base-learners may let the meta-learner leaning towards misclassifications and therefore their behavior needs to be carefully assessed through empirical evaluation. To such extent, in this paper we investigate, expand, empirically evaluate, and discuss meta-learning approaches that rely on ensembles of unsupervised algorithms to detect (zero-day) intrusions in CPSs. Our experimental comparison is conducted by means of public datasets belonging to network intrusion detection and biometric authentication systems, which are common IDSs for CPSs. Overall, we selected 21 datasets, 15 unsupervised algorithms and 9 different meta-learning approaches. Results allow discussing the applicability and suitability of meta-learning for unsupervised anomaly detection, comparing metric scores achieved by base algorithms and meta-learners. Analyses and discussion end up showing how the adoption of meta-learners significantly reduces misclassifications when detecting (zero-day) intrusions in CPSs.

Intrusion Alert Reduction Based on Unsupervised and Supervised Learning Algorithms

Security and protection of information is an ever-evolving process in the field of information security. One of the major tools of protection is the Intrusion Detection Systems (IDS). For so many years, IDS have been developed for use in computer networks, they have been widely used to detect a range of network attacks; but one of its major drawbacks is that attackers, with the evolution of time and technology make it harder for IDS systems to cope. A sub-branch of IDS-Intrusion Alert Analysis was introduced into the research system to combat these problems and help support IDS by analyzing the alert triggered by the IDS. Intrusion Alert analysis has served as a good support for IDS systems for many years but also has its own short comings which are the amount of the voluminous number of alerts produced by IDS systems. From years of research, it has been observed that majority of the alerts produced are undesirables such as duplicates, false alerts, etc., leading to huge amounts of alerts causing alert flooding. This research proposed the reduction alert by targeting these undesirable alerts through the integration of supervised and unsupervised algorithms and approach. The research first selects significant features by comparing two feature ranking techniques this targets duplicates, low priority and irrelevant alert. To achieve further reduction, the research proposed the integration of supervised and unsupervised algorithms to filter out false alerts. Based on this, an effective model was gotten which achieved 94.02% reduction rate of alerts. Making use of the dataset ISCX 2012, experiments were conducted and the model with the highest reduction rate was chosen. The model was evaluated against other experimental results and benchmarked against a related work, it also improved on the said related work.

SICEM: A Generation Approach of Band Combination for Hyperspectral Imagery Reconstitution Based on Space and Information Analyses

A band selection algorithm named space and information comprehensive evaluation model (SICEM) is proposed in this paper, which reconstitutes the hyperspectral imagery by building an optimal subset to replace the original spectrum. SICEM reduces the dimensions while keeping the vital information of an image, and these are accomplished through two phases. Specifically, the improved fast density peaks clustering (I-FDPC) algorithm is employed to pick out the scattered bands in geometric space to generate a candidate set U at first. Then, we conduct pruning in U through iterative information analysis until the target set Ω is built. In this phase, we need to calculate comprehensive information score (CIS) for every member in U after assigning weights to the amount of information (AoI) and correlation. In each iteration, the band with highest score is selected into Ω , and the ones highly related to it will be removed out of U via a threshold. Compared with the four state-of-the-art unsupervised algorithms on real-world HSI datasets (IndianP and PaviaU), we find that SICEM has strong ability to form an optimal reduced-dimension combination with low correlation and rich information and it performs well in discrete band distribution, accuracy, consistency, and stability.