Single-label Methods Research Articles

Evolutionary simultaneous under and oversampling of instances for dealing with class-imbalance datasets in multilabel problems

Multilabel classification has recently attracted great attention from the data mining research community. Multilabel classification is concerned with learning where each instance can be associated with multiple classes (or labels). Class-imbalance problems appear in any classification task when the class distribution of the instances is very different. In multilabel classification, this problem is ubiquitous, as a large percentage of labels suffer from a class-imbalanced distribution. The adaptation of single-label methods to deal with the class-imbalance problem in multilabel learning is problematic as many of their basic concepts are not easily transferred. In this paper, we propose the use of evolutionary computation to simultaneously oversample the minority class and undersample the majority class for multilabel problems. Letting the algorithm autonomously select the instances to undersample and oversample allows us to extend these two successful paradigms to the multilabel task. An extensive comparison setup of 35 datasets shows the advantages of using this approach to deal with class-imbalance datasets for multilabel problems compared with previously published methods as well as the basic classification algorithms with the original datasets.

Quantitative determination of rhamnolipid using HPLC-UV through carboxyl labeling.

Rhamnolipid, as a low-toxic, biodegradable and environmentally friendly biosurfactant, has broad application prospects in many industries. However, the quantitative determination of rhamnolipid is still a challenging task. Here, a new sensitive method for the quantitative analysis of rhamnolipid based on a simple derivatization reaction was developed. In this study, 3-[3'-(l-rhamnopyranosyloxy) decanoyloxy] decanoic acid (Rha-C10-C10) and 3-[3'-(2'-O-α-l-rhamnopyranosyloxy) decanoyloxy] decanoic acid (Rha-Rha-C10-C10) were utilized as the representative rhamnolipids. Liquid chromatography-mass spectrometry and high-performance liquid chromatography-ultra violet results showed that these two compounds were successfully labeled with 1 N1-(4-nitrophenyl)-1,2-ethylenediamine. There was an excellent linear relationship between rhamnolipid concentration and peak area of labeled rhamnolipid. The detection limits of the Rha-C10-C10 and Rha-Rha-C10-C10 were 0.018mg/L (36nmol/L) and 0.014mg/L (22nmol/L), respectively. The established amidation method was suitable for the accurate analysis of rhamnolipids in the biotechnological process. The method had good reproducibility with the relative standard deviation of 0.96% and 0.79%, respectively, and sufficient accuracy with a recovery of 96%-100%. This method was applied to quantitative analysis of 10 rhamnolipid homologs metabolized by Pseudomonas aeruginosa LJ-8. The single labeling method was used for the quantitative analysis of multiple components, which provided an effective method for the quality evaluation of other glycolipids with carboxyl groups.

ML-k’sNN: Label Dependent k Values for Multi-Label k-Nearest Neighbor Rule

Multi-label classification as a data mining task has recently attracted increasing interest from researchers. Many current data mining applications address problems with instances that belong to more than one category. These problems require the development of new, efficient methods. Multi-label k-nearest neighbors rule, ML-kNN, is among the best-performing methods for multi-label problems. Current methods use a unique k value for all labels, as in the single-label method. However, the distributions of the labels are frequently very different. In such scenarios, a unique k value for the labels might be suboptimal. In this paper, we propose a novel approach in which each label is predicted with a different value of k. Obtaining the best k for each label is stated as an optimization problem. Three different algorithms are proposed for this task, depending on which multi-label metric is the target of our optimization process. In a large set of 40 real-world multi-label problems, our approach improves the results of two different tested ML-kNN implementations.

Non-Co-Occurrence Enhanced Multi-Label Cross-Modal Hashing Retrieval Based on Graph Convolutional Network

Supervised cross-modal retrieval has significant advantages in retrieval efficiency and storage cost. In the field of hashing retrieval, existing supervised methods are divided into single-label and multi-label methods. For the single-label method, simply using a single label to measure the semantic relevance between instances will cause an error in supervision information. However, the existing multi-label hashing methods also have some problems. For example, only considering the co-occurrence of multiple labels among instances may not accurately reflect their similarity. At the same time, in the previous methods, the text modality processing did not reach the fine level of image modality, resulting in insufficient use of text information. We proposed Non-co-occurrence enhanced Multi-label cross-modal hashing retrieval based on Graph Convolutional Network (MHGCN) to address these issues. Firstly, we introduced a multi-label non-co-occurrence similarity measurement method, which adds multi-label non-co-occurrence information among instances in the multi-label similarity measurement to measure the differences between instances; Secondly, we used Graph Convolutional Networks (GCNS) to process the information on text modality; Thirdly, we introduced the memory mechanism to restrict the difference of hash code learning. Many experiments on three commonly used benchmark datasets show that the proposed Non-co-occurrence enhanced Multi-label cross-modal hashing retrieval based on Graph Convolutional Network (MHGCN) method has excellent performance.

Analysis and Detection against Network Attacks in the Overlapping Phenomenon of Behavior Attribute

The proliferation of network attacks poses a significant threat. Researchers propose datasets for network attacks to support research in related fields. Then, many attack detection methods based on these datasets are proposed. These detection methods, whether two-classification or multi-classification, belong to single-label learning, i.e., only one label is given to each sample. However, we discover that there is a noteworthy phenomenon of behavior attribute overlap between attacks, The presentation of this phenomenon in a dataset is that there are multiple samples with the same features but different labels. In this paper, we verify the phenomenon in well-known datasets(UNSW-NB15, CCCS-CIC-AndMal-2020) and re-label these data. In addition, detecting network attacks in a multi-label manner can obtain more information, providing support for tracing the attack source and building IDS. Therefore, we propose a multi-label detection model based on deep learning, MLD-Model, in which Wasserstein-Generative-Adversarial-Network-with-Gradient-Penalty (WGAN-GP) with improved loss performs data enhancement to alleviate the class imbalance problem, and Auto-Encoder (AE) performs classifier parameter pre-training. Experimental results demonstrate that MLD-Model can achieve excellent classification performance. It can achieve F1=80.06% in UNSW-NB15 and F1=83.63% in CCCS-CIC-AndMal-2020. Especially, MLD-Model is 5.99%∼7.97% higher in F1 compared with the related single-label methods.

PSIONplusm Server for Accurate Multi-Label Prediction of Ion Channels and Their Types.

Computational prediction of ion channels facilitates the identification of putative ion channels from protein sequences. Several predictors of ion channels and their types were developed in the last quindecennial. While they offer reasonably accurate predictions, they also suffer a few shortcomings including lack of availability, parallel prediction mode, single-label prediction (inability to predict multiple channel subtypes), and incomplete scope (inability to predict subtypes of the voltage-gated channels). We developed a first-of-its-kind PSIONplusm method that performs sequential multi-label prediction of ion channels and their subtypes for both voltage-gated and ligand-gated channels. PSIONplusm sequentially combines the outputs produced by three support vector machine-based models from the PSIONplus predictor and is available as a webserver. Empirical tests show that PSIONplusm outperforms current methods for the multi-label prediction of the ion channel subtypes. This includes the existing single-label methods that are available to the users, a naïve multi-label predictor that combines results produced by multiple single-label methods, and methods that make predictions based on sequence alignment and domain annotations. We also found that the current methods (including PSIONplusm) fail to accurately predict a few of the least frequently occurring ion channel subtypes. Thus, new predictors should be developed when a larger quantity of annotated ion channels will be available to train predictive models.

Classification on grade, price, and region with multi-label and multi-target methods in wineinformatics

Classifying wine according to their grade, price, and region of origin is a multi-label and multi-target problem in wine-informatics. Using wine reviews as the attributes, we compare several different multi-label/multitarget methods to the single-label method where each label is treated independently. We explore both single-label and multi-label approaches for a two-class problem for each of the labels and we explore both single-label and multi-target approaches for a four-class problem on two of the three labels, with the third label remaining a two-class problem. In terms of per-label accuracy, the single-label method has the best performance, although some multi-label methods approach the performance of single-label. However, multi-label/multi-target metrics approaches do exceed the performance of the single-label method.

ML2S-SVM: multi-label least-squares support vector machine classifiers

PurposeImage classification is becoming a supporting technology in several image-processing tasks. Due to rich semantic information contained in the images, it is very popular for an image to have several labels or tags. This paper aims to develop a novel multi-label classification approach with superior performance.Design/methodology/approachMany multi-label classification problems share two main characteristics: label correlations and label imbalance. However, most of current methods are devoted to either model label relationship or to only deal with unbalanced problem with traditional single-label methods. In this paper, multi-label classification problem is regarded as an unbalanced multi-task learning problem. Multi-task least-squares support vector machine (MTLS-SVM) is generalized for this problem, renamed as multi-label LS-SVM (ML2S-SVM).FindingsExperimental results on the emotions, scene, yeast and bibtex data sets indicate that the ML2S-SVM is competitive with respect to the state-of-the-art methods in terms of Hamming loss and instance-based F1 score. The values of resulting parameters largely influence the performance of ML2S-SVM, so it is necessary for users to identify proper parameters in advance.Originality/valueOn the basis of MTLS-SVM, a novel multi-label classification approach, ML2S-SVM, is put forward. This method can overcome the unbalanced problem but also explicitly models arbitrary order correlations among labels by allowing multiple labels to share a subspace. In addition, the multi-label classification approach has a wider range of applications. That is to say, it is not limited to the field of image classification.

Scikit-multilearn: A scikit-based Python environment for performing multi-label classification

The scikit-multilearn is a Python library for performing multi-label classification. It is compatible with the scikit-learn and scipy ecosystems and uses sparse matrices for all internal operations; provides native Python implementations of popular multi-label classification methods alongside a novel framework for label space partitioning and division and includes modern algorithm adaptation methods, network-based label space division approaches, which extracts label dependency information and multi-label embedding classifiers. The library provides Python wrapped access to the extensive multi-label method stack from Java libraries and makes it possible to extend deep learning single-label methods for multilabel tasks. The library allows multi-label stratification and data set management. The implementation is more efficient in problem transformation than other established libraries, has good test coverage and follows PEP8. Source code and documentation can be downloaded from http://scikit.ml and also via pip. The project is BSD-licensed.

Single-label time-resolved luminescence assay for estrogen receptor–ligand binding

Homogeneous luminescence-based microplate assays are desirable in high-throughput screening of new nuclear receptor regulators. Time-resolved fluorescence resonance energy transfer (TR–FRET) assays provide high sensitivity due to low background signal. The TR–FRET concept requires labeling of both ligand and receptor, making the assay format and its development relatively expensive and complex compared with single-label methods. To overcome the limitations of the multilabel methods, we have developed a single-label method for estrogen receptor (ER)–ligand binding based on quenching resonance energy transfer (QRET), where estradiol labeled with luminescent europium(III) chelate (Eu–E 2) is quenched using soluble quencher molecules. The luminescence signal of Eu–E 2 on binding to full-length ER is protected from quenching while increasing competitor concentrations displace Eu–E 2 from the receptor, reducing the signal. The QRET method was paralleled with a commercial fluorescence polarization (FP) assay. The measured signal-to-background (S/B) values for estradiol, estrone, fulvestrant, and tamoxifen obtained for the QRET assay (5.8–9.2) were clearly higher than the S/B values for the FP assay (1.3–1.5). A K d value of 30 nM was calculated for binding of Eu–E 2 to ER from a saturation binding isotherm. The QRET method provides an attractive new single-label assay format for nuclear receptor ligand screening.

Quantitative Serum Proteomics Using Dual Stable Isotope Coding and nano LC-MS/MSMS

Stable isotope coding technique in combination with mass spectrometry has emerged as a powerful tool to accurately identify and differentially quantify proteins within complex protein mixtures. We present a novel methodology to increase the yield of quantified proteins while maintaining a high stable-isotopic labeling efficacy. With this approach, intact proteins in complex biological sample such as sera are labeled with the designated dual stable isotope coding (DSIC) systems. In brief, intact proteins are coded sequentially with acrylamide to label Cysteine residues (Cys) and with succinic anhydride to label Lysine residues (Lys). Protein samples coded with this dual stable isotope are subjected to an online 2D-HPLC fractionation. The resolved protein fractions are individually digested with trypsin and analyzed with nano LC-MS/MSMS. Our results show that the DSIC labeling efficiency is 100% for Cysteine (Cys) labeled with acrylamide and 98% for Lysine (Lys) labeled with succinic anhydride. A comparative analysis of DSIC labeling and single labeling of Cysteine residues was made. Analysis of an entire anion-exchange chromatography subfraction of sera yielded 165 identified proteins (criteria: error rate <5% and unique peptides >or=2), 104 of which were quantified using the single labeling method (i.e., Cysteine acrylamide labeling only). In contrast, using same criteria for identification, a total 185 proteins were identified and 174 proteins were quantified using the DSIC labeling technique.

Effect of Lactulose on Calcium and Magnesium Absorption: A Study Using Stable Isotopes in Adult Men

To evaluate the effect of lactulose on calcium (Ca) and magnesium (Mg) absorption, we performed a clinical trial with a double-blind, randomized, crossover design in 24 healthy adult male volunteers. The absorptions of Ca and Mg were evaluated by a single-labeling method using stable isotopes. The test foods, containing lactulose at a dose of 0 g (placebo), 2 g (low-dose), or 4 g (high-dose) together with 300 mg of Ca containing 20 mg of 44Ca, and 150 mg of Mg containing 28 mg of 25Mg, were administered orally. Urine samples were collected for 8 h after the ingestion of the test food. The ratios of stable isotopes in urine (44Ca/40Ca and 25Mg/24Mg) were measured by ICP-MS (inductively coupled plasma-mass spectrometry). The urinary stable-isotopes ratios (44Ca/40Ca and 25Mg/24Mg) increased with lactulose dosage. Significant differences were observed in the Ca ratio between placebo and high-dose lactulose (p<0.01), and in the Mg ratio between placebo and low-dose lactulose and between placebo and high-dose lactulose (p<0.01). Lactulose ingestion did not change the levels of bone-resorption markers (type I collagen cross-linked N-telopeptide and deoxypyridinoline) in urine. The test foods did not cause any side effects. This study demonstrates that lactulose enhances the absorptions of Ca and Mg in adult men.

Synthesis of a Terbium Fluorescent Chelate and Its Application to Time-Resolved Fluoroimmunoassay

A new nonadentate ligand, N, N, N1, N1-[2,6-bis(3'-aminomethyl-1'-pyrazolyl)-4-phenylpyridine]tetrakis(acetic acid) (BPTA) for a Tb3+ fluorescent complex was synthesized. The Tb3+ complex is strongly fluorescent, having a large fluorescence quantum yield of 1.00 and very long fluorescence lifetime of 2.681 ms in 0.05 M borate buffer of pH 9.1. Streptavidin (SA) was labeled with BPTA by using its succinimidyl monoester, and the BPTA-Tb(3+)-labeled SA was used in sandwich-type time-resolved fluoroimmunoassay (TR-FIA) of alpha-fetoprotein (AFP) and carcinoembryonic antigen (CEA) in human sera. The Tb(3+)-labeled SA was also used in competitive-type TR-FIA of bensulfuron-methyl (BSM) in water. The detection limits of these assays are 42 pg/mL for AFP, 70 pg/mL for CEA, and 0.4 ng/mL for BSM. In addition, a new simultaneous measurement method for AFP and CEA in a human serum sample was developed by using 4,4'-bis(1",1",2",2",3",3" -heptafluoro-4",6"-hexanedion-6"-ly)chlorosulfo- o-terphenyl (BHHCT)-Eu(3+)-labeled anti-AFP antibody, biotinylated anti-CEA antibody, and BPTA-Tb(3+)-labeled SA. The concentrations of AFP and CEA in 39 human serum samples were determined, and the results were compared with those of the independently determined AFP and CEA by TR-FIA with a single-label method. A good correlation was obtained with the correlation coefficients of 0.991 for AFP and 0.994 for CEA.

Fluorescence Polarization and Anisotropy in High Throughput Screening: Perspectives and Primer

Fluorescence polarization and anisotropy are two nearly equivalent techniques that have together, over the past 5 years, achieved wide use in high throughput screening in drug discovery. These are single-label methods that can be used to construct homogeneous assays that are fast, sensitive, and resistant to some significant interferences. Moreover, the assays are relatively inexpensive. This review surveys the peer-reviewed literature on the subject and explores some of the fundamental issues that bear on assay performance.

A new apheresis procedure for the preparation of high-quality red cells and plasma.

Multicomponent apheresis is an alternative way of preparing blood components that avoids the delay between collection and separation seen with standard whole-blood techniques. An apheresis device has been modified to facilitate the combined collection of a unit (250 mL) of red cells (RBCs) and a high-volume unit (475 mL) of plasma. The procedure, using 8-percent ACD-A, has been tested in two European blood centers. Each center performed 20 procedures for in vitro evaluation of collected RBCs and plasma and 10 procedures for evaluation of in vivo RBC recovery. All RBCs were white cell reduced by filtration. One-half of the RBC units were stored in the additive solution Adsol and one-half in another such solution (Erythro-Sol). The target volumes of RBCs and plasma were obtained in 27 minutes (range, 20-44 min) by using three to six cycles in a single-needle procedure. Saline (275 mL) was used to replace fluid volume withdrawn in excess of standard whole-blood donation. No side effects occurred, with the exception of minor signs of hypocalcemia. RBC ATP was well maintained (>65% at Day 42) during storage; 2,3-DPG was less well maintained, with virtually none remaining at Day 21 in either Adsol or Erythro-Sol. The RBC in vivo recoveries, after 42 days of storage at 4+/-2 degrees C determined by the single-label method, were 86.7+/-7.2 percent (Erythro-Sol) and 84.4+/-8.1 percent (Adsol). Mean plasma factor VIII levels were >100 percent in all test groups. A novel automated technique for the simultaneous collection and preparation of RBCs and plasma has been evaluated. The apheresis procedure was acceptable and well tolerated by donors, and it resulted in high-quality blood components. Further optimization of the system should yield a practicable component suitable for routine use in blood banks.

Studies with nonradioisotopic sodium chromate

A recently developed nonradioisotopic 52Cr technique was used to measure either red cell volume or posttransfusion recovery of stored red cells. The experimental method uses Zeeman electrothermal atomic absorption spectrophotometry to measure red cell chromium. Results from the 52Cr method were compared with those from 51Cr single-label and 125I-albumin/51Cr double-label procedures using 49-day AS-1 red cell concentrates drawn and prepared according to standard procedures. In the first group of five donors, red cell volume was estimated concurrently with both 52Cr-labeled fresh red cells and 125I-albumin. The latter measured plasma volume from which red cell volume was estimated on the basis of the hematocrit (125I red cell volume). 51Cr-labeled stored red cells were transfused to measure posttransfusion recoveries. The correlation between 52Cr and 125I red cell volumes was significant (r = 0.68, p less than 0.01), and, in this group, the differences were not significant (p less than 0.05). Twenty-four-hour posttransfusion recoveries of 51Cr-labeled stored red cells averaged 66 +/- 5 percent when measured with the 125I/51Cr technique and 69 +/- 8 percent when measured with the 52Cr/51Cr method. In the second group of five donors, red cell volume was estimated by the 125I-albumin technique, and the posttransfusion recovery of stored red cells was quantitated by 51Cr- and 52Cr-labeled stored cells simultaneously. In this group, posttransfusion recoveries with 125I/51Cr averaged 73 +/- 7 percent; with 125I/52Cr, they averaged 75 +/- 10 percent. Using the single-label method of calculation, recoveries averaged 76 +/- 7 and 75 +/- 10 percent for the 51Cr and 52Cr methods, respectively.(ABSTRACT TRUNCATED AT 250 WORDS)

Estimating Bacterial Production in Marine Waters from the Simultaneous Incorporation of Thymidine and Leucine

We examined the simultaneous incorporation of [H]thymidine and [C]leucine to obtain two independent indices of bacterial production (DNA and protein syntheses) in a single incubation. Incorporation rates of leucine estimated by the dual-label method were generally higher than those obtained by the single-label method, but the differences were small (dual/single = 1.1 +/- 0.2 [mean +/- standard deviation]) and were probably due to the presence of labeled leucyl-tRNA in the cold trichloroacetic acid-insoluble fraction. There were no significant differences in thymidine incorporation between dual- and single-label incubations (dual/ single = 1.03 +/- 0.13). Addition of the two substrates in relatively large amounts (25 nM) did not apparently increase bacterial activity during short incubations (<5 h). With the dual-label method we found that thymidine and leucine incorporation rates covaried over depth profiles of the Chesapeake Bay. Estimates of bacterial production based on thymidine and leucine differed by less than 25%. Although the need for appropriate conversion factors has not been eliminated, the dual-label approach can be used to examine the variation in bacterial production while ensuring that the observed variation in incorporation rates is due to real changes in bacterial production rather than changes in conversion factors or introduction of other artifacts.

Determination of fetal, placental and neonatal blood volumes in the sheep.

A sheep preparation was developed for studying fetoplacental and neonatal blood volumes using isotope dilution techniques. Fetoplacental blood volume was measured in 46 fetal lambs (gestational age 112 to 146 days) using radioactive iodinated human serum albumin ( 125 I-RISA). In 28 of the 46 animals, blood volume was also computed using 125 I-RISA and maternal red cells labeled with radiochromium ( 51 Cr). Blood volume averaged 156.1 ml/kg of fetal weight with the single-label method and 134.7 ml/kg with the double-label method. Blood volumes measured in 17 neonatal lambs with the single-label technique averaged 110.0 ml/kg, and with the double-label technique in 10 animals averaged 104.0 ml/kg. Placental blood volume, determined by subtracting the neonatal blood volume from the fetoplacental blood volume, although varying from animal to animal, was independent of fetal weight and remained relatively constant during the last fourth of gestation.