Multiple Classes Research Articles

Fast Yet Effective Machine Unlearning.

Unlearning the data observed during the training of a machine learning (ML) model is an important task that can play a pivotal role in fortifying the privacy and security of ML-based applications. This article raises the following questions: 1) can we unlearn a single or multiple class(es) of data from an ML model without looking at the full training data even once? and 2) can we make the process of unlearning fast and scalable to large datasets, and generalize it to different deep networks? We introduce a novel machine unlearning framework with error-maximizing noise generation and impair-repair based weight manipulation that offers an efficient solution to the above questions. An error-maximizing noise matrix is learned for the class to be unlearned using the original model. The noise matrix is used to manipulate the model weights to unlearn the targeted class of data. We introduce impair and repair steps for a controlled manipulation of the network weights. In the impair step, the noise matrix along with a very high learning rate is used to induce sharp unlearning in the model. Thereafter, the repair step is used to regain the overall performance. With very few update steps, we show excellent unlearning while substantially retaining the overall model accuracy. Unlearning multiple classes requires a similar number of update steps as for a single class, making our approach scalable to large problems. Our method is quite efficient in comparison to the existing methods, works for multiclass unlearning, does not put any constraints on the original optimization mechanism or network design, and works well in both small and large-scale vision tasks. This work is an important step toward fast and easy implementation of unlearning in deep networks. Source code: https://github.com/vikram2000b/Fast-Machine-Unlearning.

Large-scale characterization of drug mechanism of action using proteome-wide thermal shift assays.

In response to an ever-increasing demand of new small molecules therapeutics, numerous chemical and genetic tools have been developed to interrogate compound mechanism of action. Owing to its ability to approximate compound-dependent changes in thermal stability, the proteome-wide thermal shift assay has emerged as a powerful tool in this arsenal. The most recent iterations have drastically improved the overall efficiency of these assays, providing an opportunity to screen compounds at a previously unprecedented rate. Taking advantage of this advance, we quantified more than one million thermal stability measurements in response to multiple classes of therapeutic and tool compounds (96 compounds in living cells and 70 compounds in lysates). When interrogating the dataset as a whole, approximately 80% of compounds (with quantifiable targets) caused a significant change in the thermal stability of an annotated target. There was also a wealth of evidence portending off-target engagement despite the extensive use of the compounds in the laboratory and/or clinic. Finally, the combined application of cell- and lysate-based assays, aided in the classification of primary (direct ligand binding) and secondary (indirect) changes in thermal stability. Overall, this study highlights the value of these assays in the drug development process by affording an unbiased and reliable assessment of compound mechanism of action.

Multiplicity dependence of the pT -spectra for identified particles and its relationship with partonic entropy

We investigate the multiplicity dependence of the transverse momentum pT spectra of hadrons produced in high-energy collisions. We propose that the partonic distribution be parametrized by its nonextensive entropy and the parton saturation scale Qs(x). These two variables can be identified from the produced charged hadron distributions and provide important information on the gluon dynamics at the moment of interaction. From this perspective, we interpret data from different ALICE multiplicity classes at s=13 TeV and s=5.02 TeV. A multiplicity-dependent scaling function is presented and the dependence of the interaction area on multiplicity is also investigated. Published by the American Physical Society 2024

Synchronization transition in space–time chaos in the presence of quenched disorder

Synchronization of two replicas of coupled map lattices for continuous maps is known to be in the multiplicative noise universality class. We study this transition in the presence of quenched disorder in coupling. The disorder is identical in both replicas. We study one-dimensional, two-dimensional, and globally coupled logistic and tent maps. We observe a clear second-order transition with new exponents. The order parameter decays as t−δ and δ depends on the map and its parameters. The asymptotic order parameter for Δ distance from a critical point grows as Δβ with β∼δ. The quenched disorder in coupling is a relevant perturbation for the replica synchronization of coupled map lattices. The critical exponents are different from those of the multiplicative noise universality class. However, it does not depend on dimensionality in the cases studied if the transition is continuous.

Analyzing the correlation between thermal and kinematic parameters in various multiplicity classes within 7 and 13 TeV pp collisions

We investigate the transverse momentum spectra of identified particles at 7 and 13 TeV in pp collisions in the framework of the blast wave model with Tsallis statistics (TBW). Based on experimental data by ALICE Collaboration, we observe that the model describes the p T spectra well with the common Tsallis temperature (T) and flow velocity (β T ) but separate non-extensive parameters (q) for baryons and mesons. The parameter dependence on multiplicity as well as on collision energy is investigated, and a strong dependence on the former while a weak dependence on the latter is reported. The extracted parameters in this work consist of the initial temperature (T i ), the average transverse momentum (〈p T 〉), the Tsallis temperature (T), the flow velocity (β T ), and the non-extensive parameter (q). These parameters are found to increase a little with increasing energy, however, they (except the parameter q) decrease significantly with decreasing multiplicity. We observe that β T drops to zero at high multiplicity , while, T and q do not change their behavior. Furthermore, our analysis explore the correlations among different parameters, including associations with the charged particle multiplicity per unit pseudorapidity ( 〈dNch/dη〉 ). The correlations between T and β T , T and 〈dNch/dη〉 , β T and 〈dNch/dη〉 , T i and 〈p T 〉 and T i and 〈dNch/dη〉 demonstrates a positive relationship, while, the correlation between T and q − 1, and q − 1 and 〈dNch/dη〉 is negative. Finally, we implement an extra flow correction on the T parameter. Our findings reveal that the Doppler-corrected temperature parameter aligns closely with the T in scenarios with lower multiplicities. However, as the multiplicity increases, a noticeable divergence emerges between these parameters, indicating a widening separation between them.

Multi-Label Noise Robust Collaborative Learning for Remote Sensing Image Classification.

The development of accurate methods for multi-label classification (MLC) of remote sensing (RS) images is one of the most important research topics in RS. The MLC methods based on convolutional neural networks (CNNs) have shown strong performance gains in RS. However, they usually require a high number of reliable training images annotated with multiple land-cover class labels. Collecting such data is time-consuming and costly. To address this problem, the publicly available thematic products, which can include noisy labels, can be used to annotate RS images with zero-labeling cost. However, multi-label noise (which can be associated with wrong and missing label annotations) can distort the learning process of the MLC methods. To address this problem, we propose a novel multi-label noise robust collaborative learning (RCML) method to alleviate the negative effects of multi-label noise during the training phase of a CNN model. RCML identifies, ranks, and excludes noisy multi-labels in RS images based on three main modules: 1) the discrepancy module; 2) the group lasso module; and 3) the swap module. The discrepancy module ensures that the two networks learn diverse features, while producing the same predictions. The task of the group lasso module is to detect the potentially noisy labels assigned to multi-labeled training images, while the swap module is devoted to exchange the ranking information between two networks. Unlike the existing methods that make assumptions about noise distribution, our proposed RCML does not make any prior assumption about the type of noise in the training set. The experiments conducted on two multi-label RS image archives confirm the robustness of the proposed RCML under extreme multi-label noise rates. Our code is publicly available at: https://www.noisy-labels-in-rs.org.

Average-Case Analysis of Greedy Matching for Large-Scale D2D Resource Sharing

Given the proximity of many wireless users and their diversity in consuming local resources (e.g., data-plans, computation and energy resources), device-to-device (D2D) resource sharing is a promising approach towards realizing a sharing economy. This paper adopts an easy-to-implement greedy matching algorithm with distributed fashion and only sub-linear <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$O(\log n)$</tex-math></inline-formula> parallel complexity (in user number <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$n$</tex-math></inline-formula> ) for large-scale D2D sharing. Practical cases indicate that the greedy matching's average performance is far better than the worst-case approximation ratio 50% as compared to the optimum. However, there is no rigorous average-case analysis in the literature to back up such encouraging findings and this paper is the first to present such analysis for multiple representative classes of graphs. For 1D linear networks, we prove that our greedy algorithm performs better than 86.5% of the optimum. For 2D grids, though dynamic programming cannot be directly applied, we still prove this average performance ratio to be above 76%. For the more challenging Erdos-Rényi random graphs, we equivalently reduce to the asymptotic analysis of random trees and successfully prove a ratio up to 79%. Finally, we conduct experiments using real data to simulate realistic D2D networks, and show that our analytical performance measure approximates well practical cases.

Scalable SoftGroup for 3D Instance Segmentation on Point Clouds.

This paper considers a network referred to as SoftGroup for accurate and scalable 3D instance segmentation. Existing state-of-the-art methods produce hard semantic predictions followed by grouping instance segmentation results. Unfortunately, errors stemming from hard decisions propagate into the grouping, resulting in poor overlap between predicted instances and ground truth and substantial false positives. To address the abovementioned problems, SoftGroup allows each point to be associated with multiple classes to mitigate the uncertainty stemming from semantic prediction. It also suppresses false positive instances by learning to categorize them as background. Regarding scalability, the existing fast methods require computational time on the order of tens of seconds on large-scale scenes, which is unsatisfactory and far from applicable for real-time. Our finding is that the k-Nearest Neighbor ( k-NN) module, which serves as the prerequisite of grouping, introduces a computational bottleneck. SoftGroup is extended to resolve this computational bottleneck, referred to as SoftGroup++. The proposed SoftGroup++ reduces time complexity with octree k-NN and reduces search space with class-aware pyramid scaling and late devoxelization. Experimental results on various indoor and outdoor datasets demonstrate the efficacy and generality of the proposed SoftGroup and SoftGroup++. Their performances surpass the best-performing baseline by a large margin (6% ∼ 16%) in terms of AP 50. On datasets with large-scale scenes, SoftGroup++ achieves a 6× speed boost on average compared to SoftGroup. Furthermore, SoftGroup can be extended to perform object detection and panoptic segmentation with nontrivial improvements over existing methods.

Comparative Analysis of Cyanotoxins in Fishponds in Nigeria and South Africa

Over the decades, the aquaculture sector has witnessed substantial growth, contributing significantly to the nation’s economy. However, the menace of CyanoHABs threatens the sustainability of fish farming. Considering the possible hazards linked to cyanotoxins in food and water, a comparative study design between commercial fish in Nigeria and South Africa was employed to investigate cyanotoxins in the water from fishponds. Six commercial fishponds in Calabar Municipality—Nigeria and Duthuni—South Africa with varying climatic zones were selected. Water samples from the ponds were collected at intervals during different seasons (summer, winter, dry, and wet seasons) to capture climate-induced variation. Liquid chromatography–mass spectrometry (LCMS) in combination with the metabolites database was used for the identification of toxic cyanometabolites in water samples. The molecular networking approach, coupled with the Global Natural Products Social Molecular Networking (GNPS) database and CANOPUS annotation, enabled the putative identification of cyanometabolites. The resulting molecular network unveiled discernible clusters representing related molecule families, aiding in the identification of both known cyanotoxins and unfamiliar analogues. Furthermore, the molecular network revealed that water samples from different fishponds shared specific metabolites, including ethanesulfonic acid, pheophorbide A, cholic acid, phenylalanine, amyl amine, phosphocholine (PC), and sulfonic acid, despite variations in location, local climatic factors, and sampling sites. The fishponds in Nigeria showed the presence of multiple cyanotoxin classes in the dry, wet, and summer seasons in the water. Aflatoxin was identified in all sampling sites in Nigeria (N1, N2, and N3). The Duthuni, South Africa, sampling sites (P1, P2, and P3) exhibited the presence of microginins and microcystins. All the fishponds displayed a widespread occurrence of anabaenopeptins, aplysiatoxins, aflatoxin, microcolins, and marabmids during the selected summer. In conclusion, the untargeted metabolome analysis, guided by GNPS, proved highly effective in identifying both toxic and non-toxic metabolites in fishponds.

Abstract 2265: Distinct alternative splicing patterns in genetic subtypes of diffuse large B-cell lymphoma

Abstract Advances in sequencing techniques have made molecular subgrouping of Diffuse Large B-cell Lymphoma (DLBCL) based on DNA cahnges, which has led to the development of Lymphgen classification. By the way, although alternative splicing (AS) in cancer cells contributes to proliferation, invasion and metastasis, and immune escape, their role and association with DLBCL molecular subtypes are unknown. Therefore, we analyzed whole-exome sequencing (WES) and bulk RNA-Seq data paired in the public EGA dataset (EGAD00001003600) to explore the patterns of AS events that differ between Lymphgene subtypes. The WES data of 620 DLBCL samples that passed quality control were aligned to the reference using BWA-MEM, and variants were called with Mutect2. We used the information of copy number variants provided by cBioPortal. The RNA-Seq analysis was performed using the STAR-RSEM pipeline and differential splicing events were detected using rMATS to compare subtypes. Among 620 DLBCL samples, there were 19 MCDs, 95 EZBs, 22 BN2s, and 137 ST2s (50 samples with multiple genetic classes, and 291 samples unclassified). Mutations in spliceosome complex genes (ZRSR2 and SF1) were found in 14 and 27 samples, respectively. In BN2 and MCD subtypes, splicing events with significantly different exon inclusion levels were observed at a high frequency of 29 and 16 occurrences, respectively, compared to other subtypes. (P-value &amp;lt; 0.05). In contrast, retained intron is the only event observed in the EZB subtype (Table 1). Samples with mutations in ZRSR2 and SF1 showed four and two aberrant events, respectively, compared to samples without these mutations. Of note, rMATS detected an alternative 3’ splice site event in exon 6 of CD37 in the MCD. CD37 is involved in tumorigenesis in leukocytes and has been suggested as a potential biomarker in blood cancer, and its dysregulation may have influenced the poor prognosis of MCD. In summary, molecular subtypes of DLBCL had unique features in terms of alternative splicing. Functional consequences of these selective exon usages for each DLBCL subtype are being warranted. Citation Format: Sanyeowool An, Youngil Koh, Sung-Soo Yoon. Distinct alternative splicing patterns in genetic subtypes of diffuse large B-cell lymphoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2265.

NAP-seq reveals multiple classes of structured noncoding RNAs with regulatory functions

Up to 80% of the human genome produces “dark matter” RNAs, most of which are noncapped RNAs (napRNAs) that frequently act as noncoding RNAs (ncRNAs) to modulate gene expression. Here, by developing a method, NAP-seq, to globally profile the full-length sequences of napRNAs with various terminal modifications at single-nucleotide resolution, we reveal diverse classes of structured ncRNAs. We discover stably expressed linear intron RNAs (sliRNAs), a class of snoRNA-intron RNAs (snotrons), a class of RNAs embedded in miRNA spacers (misRNAs) and thousands of previously uncharacterized structured napRNAs in humans and mice. These napRNAs undergo dynamic changes in response to various stimuli and differentiation stages. Importantly, we show that a structured napRNA regulates myoblast differentiation and a napRNA DINAP interacts with dyskerin pseudouridine synthase 1 (DKC1) to promote cell proliferation by maintaining DKC1 protein stability. Our approach establishes a paradigm for discovering various classes of ncRNAs with regulatory functions.

Comparison of late-onset and non-late-onset systemic lupus erythematosus individuals in a real-world electronic health record cohort.

Objective: Late-onset systemic lupus erythematosus (LO-SLE) is defined as SLE diagnosed at age 50 years or later. Current studies on LO-SLE are small and have conflicting results.Methods: Using a large, electronic health record (EHR)-based cohort of SLE individuals, we compared demographics, disease characteristics, SLE-specific antibodies, and medication prescribing practices in LO (n = 123) vs. NLO-SLE (n = 402) individuals.Results: The median age (interquartile range) at SLE diagnosis was 60 (56-67) years for LO-SLE and 28 (20-38) years for NLO-SLE. Both groups were predominantly female (85% vs. 91%, p = 0.10). LO-SLE individuals were more likely to be White than NLO-SLE individuals (74% vs. 60%, p = 0.005) and less likely to have positive dsDNA (39% vs. 58%, p = 0.001) and RNP (17% vs. 32%, p = 0.02) with no differences in Smith, SSA, and SSB. Autoantibody positivity declined with increasing age at SLE diagnosis. LO-SLE individuals were less likely to develop SLE nephritis (9% vs. 29%, p < 0.001) and less likely to be prescribed multiple classes of SLE medications including antimalarials (90% vs. 95%, p = 0.04), azathioprine (17% vs. 31%, p = 0.002), mycophenolate mofetil (12% vs. 38%, p < 0.001), and belimumab (2% vs. 8%, p = 0.02).Conclusion: LO-SLE individuals may be less likely to fit an expected course for SLE with less frequent positive autoantibodies at diagnosis and lower rates of nephritis, even after adjusting for race. Understanding how age impacts SLE disease presentation could help reduce diagnostic delays in SLE.

High‐Throughput Colorimetric Detection and Quantification of Indoles and Pyrroloindoles for Enzymatic Activity Determination

AbstractIndoles and pyrroloindoles are structural motifs present in many biologically active natural products. Multiple classes of enzymes catalyze the transformation of indoles into pyrroloindoles via group transfer followed by intramolecular cyclization, such as peroxydases, methyltransferases, and prenyltransferases. Due to the selective introduction of a stereogenic center, these enzymes receive increasing attention as catalytic tools for the production of pharmacologically relevant compounds. Two new colorimetric assays are described in this work, which allow for the quantification of such enzymatic reactions from the perspective of the substrate and the product. For the substrates, the indole assay is based on a modified version of the Ehrlich test, with the use of light as a driving force for color formation. The pyrroloindole assay uses cerium sulfate as a reagent for the colorimetric quantification of the enzymatic products. The assays are complementary and both were successfully utilized for enzymatic activity determination of a C3‐indole methyltransferase. They can facilitate high‐throughput screening of mutant libraries, offering support for the engineering of such enzymes, but can also be used as stand‐alone methods for the detection and quantification of natural products.

Hybrid supervised and reinforcement learning for the design and optimization of nanophotonic structures.

From higher computational efficiency to enabling the discovery of novel and complex structures, deep learning has emerged as a powerful framework for the design and optimization of nanophotonic circuits and components. However, both data-driven and exploration-based machine learning strategies have limitations in their effectiveness for nanophotonic inverse design. Supervised machine learning approaches require large quantities of training data to produce high-performance models and have difficulty generalizing beyond training data given the complexity of the design space. Unsupervised and reinforcement learning-based approaches on the other hand can have very lengthy training or optimization times associated with them. Here we demonstrate a hybrid supervised learning and reinforcement learning approach to the inverse design of nanophotonic structures and show this approach can reduce training data dependence, improve the generalizability of model predictions, and significantly shorten exploratory training times. The presented strategy thus addresses several contemporary deep learning-based challenges, while opening the door for new design methodologies that leverage multiple classes of machine learning algorithms to produce more effective and practical solutions for photonic design.

Abstract B027: Repurposing ion channel drugs: the Na+ channel modifying drug DPI-201-106 significantly enhances the anti-tumor activity of multiple classes of drugs targeting DNA damage repair

Abstract B027: Repurposing ion channel drugs: the Na+ channel modifying drug DPI-201-106 significantly enhances the anti-tumor activity of multiple classes of drugs targeting DNA damage repair

Early detection of OXA-232-producing Klebsiella pneumoniae in China predating its global emergence

Antibiotic resistance is a global health issue, with Klebsiella pneumoniae (KP) posing a particular threat due to its ability to acquire resistance to multiple drug classes rapidly. OXA-232 is a carbapenemase that confers resistance to carbapenems, a class of antibiotics often used as a last resort for treating severe bacterial infections. The study reports the earliest known identification of six OXA-232-producing KP strains that were isolated in Zhejiang, China, in 2008 and 2009 within a hospital, two years prior to the first reported identification of OXA-232 in France. The four KP strains carry the OXA-232 gene and exhibit hypervirulent loci, suggesting a broader temporal and geographical spread and integration of this resistance and virulence than previously recognized with implications for public health. Global analysis of all OXA-232-bearing KP strains revealed that OXA-232-encoding plasmids are conservative, while the strains were very diverse suggesting the plasmid mediated transmission of this carbapenemase genes. Importantly, a large proportion of the OXA-232-bearing KP strains also carried virulence plasmids, in particular the recent emergence of ST15 type of KP that carried both OXA-232-encoding plasmids and hypervirulent (hv) plasmids in China since 2019, highlighting the importance of the emergence of this type of KP strains in clinical setting. The early detection and investigations of OXA-232 in these strains warrants the retrospective studies to uncover the true timeline of antibiotic resistance spread, which could provide valuable insights for shaping future strategies to tackle the global health crisis.

Epistatic pathways can drive HIV-1 escape from integrase strand transfer inhibitors.

People living with human immunodeficiency virus (HIV) receiving integrase strand transfer inhibitors (INSTIs) have been reported to experience virological failure in the absence of resistance mutations in integrase. To elucidate INSTI resistance mechanisms, we propagated HIV-1 in the presence of escalating concentrations of the INSTI dolutegravir. HIV-1 became resistant to dolutegravir by sequentially acquiring mutations in the envelope glycoprotein (Env) and the nucleocapsid protein. The selected Env mutations enhance the ability of the virus to spread via cell-cell transfer, thereby increasing the multiplicity of infection (MOI). While the selected Env mutations confer broad resistance to multiple classes of antiretrovirals, the fold resistance is ~2 logs higher for INSTIs than for other classes of drugs. We demonstrate that INSTIs are more readily overwhelmed by high MOI than other classes of antiretrovirals. Our findings advance the understanding of how HIV-1 can evolve resistance to antiretrovirals, including the potent INSTIs, in the absence of drug-target gene mutations.

G2019S selective LRRK2 kinase inhibitor abrogates mitochondrial DNA damage

Pathogenic mutations in LRRK2 cause Parkinson’s disease (PD). The G2019S variant is the most common, which results in abnormally high kinase activity. Compounds that target LRRK2 kinase activity are currently being developed and tested in clinical trials. We recently found that G2019S LRRK2 causes mitochondrial DNA (mtDNA) damage and treatment with multiple classes of LRRK2 kinase inhibitors at concentrations associated with dephosphorylation of LRRK2 reversed mtDNA damage to healthy control levels. Because maintaining the normal function of LRRK2 in heterozygous G2019S LRRK2 carriers while specifically targeting the G2019S LRRK2 activity could have an advantageous safety profile, we explored the efficacy of a G2019S mutant selective LRRK2 inhibitor to reverse mtDNA damage in G2019S LRRK2 models and patient cells relative to non-selective LRRK2 inhibitors. Potency of LRRK2 kinase inhibition by EB-42168, a G2019S mutant LRRK2 kinase inhibitor, and MLi-2, a non-selective inhibitor, was determined by measuring phosphorylation of LRRK2 at Ser935 and/or Ser1292 using quantitative western immunoblot analysis. The Mito DNADX assay, which allows for the accurate real-time quantification of mtDNA damage in a 96-well platform, was performed in parallel. We confirmed that EB-42168 selectively inhibits LRRK2 phosphorylation on G2019S LRRK2 relative to wild-type LRRK2. On the other hand, MLi-2 was equipotent for wild-type and G2019S LRRK2. Acute treatment with EB-42168 inhibited LRRK2 phosphorylation and also restored mtDNA damage to healthy control levels. We further investigated the relationship between LRRK2 kinase activity, mtDNA damage and mitophagy. Levels of mtDNA damage caused by G2019S LRRK2 were fully re-established within 2 h of a LRRK2 inhibitor wash out and recovery experiment, indicating the mtDNA damage phenotype is highly dynamic. G2019S LRRK2 mitophagy defects were not alleviated with LRRK2 kinase inhibition, suggesting that mitophagy is not mechanistically regulating LRRK2 kinase-mediated reversal of mtDNA damage in this acute timeframe. Abrogation of mtDNA damage with the mutant selective tool inhibitor EB-42168 demonstrates the potential of a precision medicine approach for LRRK2 G2019S PD. Levels of mtDNA damage may serve as a potential pharmacodynamic biomarker of altered kinase activity that could be useful for small molecule development and clinical trials.

Enhanced PUE Detection and Selfish Secondary User Detection Method in Cognitive Radio Networks

In cognitive radio networks (CRNs),cognitive radio (CR) nodes adaptively access the spectrum aiming to maximize the utilization of the scarce resource. A new security threat known as the Primary User Emulation attack raises a great challenge to CRNs. In the proposed method an innovative technique is introduced which is called as Enhanced PUE detection method and selfish secondary user detection method. In the Enhanced PUE detection, PUE detection approach can be extended to address the scenario with multiple classes of PUs that have different SAP features. The Selfish Secondary User detection method, attacks are detected by the cooperation of other legitimate neighboring Secondary Users.

TOPSIS-ACO based feature selection for multi-label classification

This paper introduces an approach for optimal feature selection (FS) in multi-label (ML) data, where each sample can be associated with multiple class labels. The complexity of the feature space is significantly increased in comparison to single-label data, making decision-making challenging. Therefore, FS plays a crucial role in ML classification. The proposed method, called ML-TOPSIS-ACO, leverages the ‘Technique of Order Preference by Similarity to Ideal Solution’ (TOPSIS) as a Multi-Criteria Decision Making (MCDM) technique and ‘Ant Colony Optimization’ (ACO) as a meta-heuristic technique. The ML-TOPSIS-ACO method operates in two phases. In the first phase, Modified TOPSIS (MTOPSIS) is used to perform feature ranking and select the most important features. These selected features are then passed to Modified-ACO for feature re-ranking, aiming to identify the best subset of features for ML classification. To evaluate the effectiveness of ML-TOPSIS-ACO, experiments were conducted on nine benchmark datasets. The experimental results demonstrate that ML-TOPSIS-ACO outperforms other existing methods, achieving a winning percentage of 83%.