Robust Label Research Articles

Multi-source adaptation learning with global and local regularization by exploiting joint kernel sparse representation

While the advantages of using information from multi-source domains for establishing an adaptation model have been widely recognized, how to boosting the robustness of the learning model has only recently received attention. For achieving a robust adaptation learning model for visual classification tasks, by exploiting multi-source adaptation regularization joint kernel sparse representation (ARJKSR), we propose a novel Multi-source Adaptation learning method with Global and Local Regularization (MAGLR), which aims to minimize a cost function that properly trades off the global and local learning costs so as to utilize both the local and global discriminative information of target domain. Specifically, we first learn the optimal ARJKSR of domain datasets, which jointly represents the datasets by a sparse linear combination of a self-expressive dictionary in some optimal reproduced kernel Hilbert space (RKHS) recovered by minimizing the inter-domain distribution discrepancy, whilst constraining the observations from domains to share their sparse representations. Thus, we simultaneously take into account correlations as well as coupling information among multiple source domains. Based on the ARJKSR, we then derive a robust label prediction matrix for unlabeled instances from the target domain based on the classical graph-based semi-supervised learning diagram, into which global and local regularization are incorporated. Furthermore, a multi-kernel regression way is exploited to extend MAGLR to out-of-sample data. The validity of our method is examined by several real-world visual classification tasks. Experimental results demonstrate the superiority of our method.

Multi-source adaptation joint kernel sparse representation for visual classification

Most of the existing domain adaptation learning (DAL) methods relies on a single source domain to learn a classifier with well-generalized performance for the target domain of interest, which may lead to the so-called negative transfer problem. To this end, many multi-source adaptation methods have been proposed. While the advantages of using multi-source domains of information for establishing an adaptation model have been widely recognized, how to boost the robustness of the computational model for multi-source adaptation learning has only recently received attention. To address this issue for achieving enhanced performance, we propose in this paper a novel algorithm called multi-source Adaptation Regularization Joint Kernel Sparse Representation (ARJKSR) for robust visual classification problems. Specifically, ARJKSR jointly represents target dataset by a sparse linear combination of training data of each source domain in some optimal Reproduced Kernel Hilbert Space (RKHS), recovered by simultaneously minimizing the inter-domain distribution discrepancy and maximizing the local consistency, whilst constraining the observations from both target and source domains to share their sparse representations. The optimization problem of ARJKSR can be solved using an efficient alternative direction method. Under the framework ARJKSR, we further learn a robust label prediction matrix for the unlabeled instances of target domain based on the classical graph-based semi-supervised learning (GSSL) diagram, into which multiple Laplacian graphs constructed with the ARJKSR are incorporated. The validity of our method is examined by several visual classification problems. Results demonstrate the superiority of our method in comparison to several state-of-the-arts.

Robust Face Label Graph Matching for Movie Character Identification

a successful application in the field of image analysis, face recognition is seeking most of the attention during the past few years. It is a difficult problem because of the huge variation in the appearance of each character and the weakness and ambiguity of available annotation. Robust Face Name Graph Matching for Movie Character Identification is used to detect the faces of the movie characters and the system designed is taking the least time to detect the face. Through the proposed system, we will be able to do it in a one minute process.

Violacein as a genetically-controlled, enzymatically amplified and photobleaching-resistant chromophore for optoacoustic bacterial imaging.

There is growing interest in genetically expressed reporters for in vivo studies of bacterial colonization in the context of infectious disease research, studies of the bacterial microbiome or cancer imaging and treatment. To empower non-invasive high-resolution bacterial tracking with deep tissue penetration, we herein use the genetically controlled biosynthesis of the deep-purple pigment Violacein as a photobleaching-resistant chromophore label for in vivo optoacoustic (photoacoustic) imaging in the near-infrared range. We demonstrate that Violacein-producing bacteria can be imaged with high contrast-to-noise in strongly vascularized xenografted murine tumors and further observe that Violacein shows anti-tumoral activity. Our experiments thus identify Violacein as a robust bacterial label for non-invasive optoacoustic imaging with high potential for basic research and future theranostic applications in bacterial tumor targeting.

Designer labels for plant metabolism: statistical design of isotope labeling experiments for improved quantification of flux in complex plant metabolic networks

Metabolic fluxes are powerful indicators of cell physiology and can be estimated by isotope-assisted metabolic flux analysis (MFA). The complexity of the compartmented metabolic networks of plants has constrained the application of isotope-assisted MFA to them, principally because of poor identifiability of fluxes from the measured isotope labeling patterns. However, flux identifiability can be significantly improved by a priori design of isotope labeling experiments (ILEs). This computational design involves evaluating the effect of different isotope label and isotopomer measurement combinations on flux identifiability, and thereby identifying optimal labels and measurements toward evaluating the fluxes of interest with the highest confidence. This article reports ILE designs for two major, compartmented plant metabolic pathways - the pentose phosphate pathway (PPP) and γ-aminobutyric acid (GABA) shunt. Together, these pathways represent common motifs in plant metabolism including duplication of pathways in different subcellular compartments, reversible reactions and cyclic carbon flow. To compare various ILE designs, we employed statistical A- and D-optimality criteria. Our computations showed that 1,2-(13)C Glc is a powerful and robust label for the plant PPPs, given currently popular isotopomer measurement techniques (single quadrupole mass spectrometry [MS] and 2-D nuclear magnetic resonance [NMR]). Further analysis revealed that this label can estimate several PPP fluxes better than the popular label 1-(13)C Glc. Furthermore, the concurrent measurement of the isotopomers of hexose and pentose moieties synthesized exclusively in the cytosol or the plastid compartments (measurable through intracellular glucose or sucrose, starch, RNA ribose and histidine) considerably improves the identifiability of PPP fluxes in the individual compartments. Additionally, MS-derived isotopomer measurements outperform NMR-derived measurements in identifying PPP fluxes. The potency of 1,2-(13)C Glc can be improved substantially by combining it with other labels (e.g. 3-(13)C Glc, 1-(13)C Glc and U-(13)C Glc) in parallel ILEs. For the GABA shunt, we calculated that 100% 2-(13)C Ala and 100% U-(13)C Gln constitute the best labels. We anticipate that the ILE designs presented in this article can enhance the quality of flux estimates in these two complex plant pathways. In the future, these ILE designs can be further improved by leveraging recent analytical and computational developments in isotope-assisted MFA.

Lack of robust LGN label following transneuronal rabies virus injections into macaque area V4

In primates, retinal inputs are relayed through the magno- and parvocells of the lateral geniculate nucleus (LGN) indirectly to extrastriate visual cortex. The most direct pathway identified to the extrastriate cortex is a disynaptic one that provides robust magno- and parvocellular inputs to the middle temporal area (MT). The inclusion of parvocells in this projection is somewhat surprising because of their importance for color and form vision, whereas MT is more strictly tuned to velocity. This raises the question of whether areas more involved in color and form processing, such as V4, receive similar projections. We report here on experiments that use rabies virus injections into V4 to retrogradely label mono- and disynaptic inputs. We find only a small number of labeled neurons in the LGN in a pattern consistent with monosynaptic labeling of koniocells, rather than disynaptic labeling of magno- and parvocells. The lack of robust magno- and parvocellular label was not due to ineffective viral transport because in the same cases we find hundreds of neurons labeled in the thalamic reticular nucleus, a structure that can only be labeled disynaptically from the cortex. We also find a complete absence of neurons labeled in V1, but thousands in adjacent areas V2 and V3. This result helps explain the absence of labeled magno- and parvocells in the LGN because disynaptic transport from an extrastriate visual area should require a relay through V1. Taken together, these results suggest that ascending magno/parvocellular inputs to V4 are more hierarchically organized than the relatively direct inputs to MT.

Magnetic Relaxation Detector for Microbead Labels.

A compact and robust magnetic label detector for biomedical assays is implemented in 0.18-μm CMOS. Detection relies on the magnetic relaxation signature of a microbead label for improved tolerance to environmental variations and relaxed dynamic range requirement, eliminating the need for baseline calibration and reference sensors. The device includes embedded electromagnets to eliminate external magnets and reduce power dissipation. Correlated double sampling combined with offset servo loops and magnetic field modulation, suppresses the detector offset to sub-μT. Single 4.5-μm magnetic beads are detected in 16 ms with a probability of error <0.1%.

Facile synthesis of water-soluble, highly fluorescent graphene quantum dots as a robust biological label for stem cells

We present a facile electrochemical method for synthesizing uniform sized graphene quantum dots (GQDs) with a strong yellow emission at 14% quantum yield. This approach has enabled a large-scale production of aqueous GQD solution without the need for polymeric or surfactant stabilizers. The structure and emission mechanism of the GQDs have been studied by combining extensive characterization techniques, rigorous control experiments and theoretical calculations. We further demonstrate the distinctive advantages of such GQDs for direct and efficient stem cell labeling, opening up great opportunities for their bio-medical applications.

A trigeminoreticular pathway: implications in pain.

Neurons in the caudalmost ventrolateral medulla (cmVLM) respond to noxious stimulation. We previously have shown most efferent projections from this locus project to areas implicated either in the processing or modulation of pain. Here we show the cmVLM of the rat receives projections from superficial laminae of the medullary dorsal horn (MDH) and has neurons activated with capsaicin injections into the temporalis muscle. Injections of either biotinylated dextran amine (BDA) into the MDH or fluorogold (FG)/fluorescent microbeads into the cmVLM showed projections from lamina I and II of the MDH to the cmVLM. Morphometric analysis showed the retrogradely-labeled neurons were small (area 88.7 µm2±3.4) and mostly fusiform in shape. Injections (20–50 µl) of 0.5% capsaicin into the temporalis muscle and subsequent immunohistochemistry for c-Fos showed nuclei labeled in the dorsomedial trigeminocervical complex (TCC), the cmVLM, the lateral medulla, and the internal lateral subnucleus of the parabrachial complex (PBil). Additional labeling with c-Fos was seen in the subnucleus interpolaris of the spinal trigeminal nucleus, the rostral ventrolateral medulla, the superior salivatory nucleus, the rostral ventromedial medulla, and the A1, A5, A7 and subcoeruleus catecholamine areas. Injections of FG into the PBil produced robust label in the lateral medulla and cmVLM while injections of BDA into the lateral medulla showed projections to the PBil. Immunohistochemical experiments to antibodies against substance P, the substance P receptor (NK1), calcitonin gene regulating peptide, leucine enkephalin, VRL1 (TPRV2) receptors and neuropeptide Y showed that these peptides/receptors densely stained the cmVLM. We suggest the MDH- cmVLM projection is important for pain from head and neck areas. We offer a potential new pathway for regulating deep pain via the neurons of the TCC, the cmVLM, the lateral medulla, and the PBil and propose these areas compose a trigeminoreticular pathway, possibly the trigeminal homologue of the spinoreticulothalamic pathway.

Robust statistical label fusion through COnsensus Level, Labeler Accuracy, and Truth Estimation (COLLATE).

Segmentation and delineation of structures of interest in medical images is paramount to quantifying and characterizing structural, morphological, and functional correlations with clinically relevant conditions. The established gold standard for performing segmentation has been manual voxel-by-voxel labeling by a neuroanatomist expert. This process can be extremely time consuming, resource intensive and fraught with high inter-observer variability. Hence, studies involving characterizations of novel structures or appearances have been limited in scope (numbers of subjects), scale (extent of regions assessed), and statistical power. Statistical methods to fuse data sets from several different sources (e.g., multiple human observers) have been proposed to simultaneously estimate both rater performance and the ground truth labels. However, with empirical datasets, statistical fusion has been observed to result in visually inconsistent findings. So, despite the ease and elegance of a statistical approach, single observers and/or direct voting are often used in practice. Hence, rater performance is not systematically quantified and exploited during label estimation. To date, statistical fusion methods have relied on characterizations of rater performance that do not intrinsically include spatially varying models of rater performance. Herein, we present a novel, robust statistical label fusion algorithm to estimate and account for spatially varying performance. This algorithm, COnsensus Level, Labeler Accuracy and Truth Estimation (COLLATE), is based on the simple idea that some regions of an image are difficult to label (e.g., confusion regions: boundaries or low contrast areas) while other regions are intrinsically obvious (e.g., consensus regions: centers of large regions or high contrast edges). Unlike its predecessors, COLLATE estimates the consensus level of each voxel and estimates differing models of observer behavior in each region. We show that COLLATE provides significant improvement in label accuracy and rater assessment over previous fusion methods in both simulated and empirical datasets.

Bromodeoxyuridine infused into the cerebral ventricle of adult mice labels nigral neurons under physiological conditions—A method to detect newborn nerve cells in regions with a low rate of neurogenesis

Methodological differences may explain discrepancies in adult mammalian neurogenesis in the brain outside the widely accepted neurogenic regions, i.e. hippocampus and olfactory bulb/subventricular zone. Here, we describe a method to dissolve and administer bromodeoxyuridine (BrdU) at high concentrations (150mg/mL) into the adult mouse brain to demonstrate neuronal incorporation of this thymidine analogue in CNS regions with a low rate of neurogenesis. The dosage and duration of BrdU appear critical, since exposure to low doses did not result in a robust label using this marker for proliferation [Zhao et al., 2003. Proc Natl Acad Sci USA 100; 7925]. Mice (five per cage in an enriched environment) received BrdU in the right cerebral ventricle from a subcutaneous osmotic pump (0.9mg/day, 3 weeks, infused at 0.25μL/h). To avoid labelling of cells with a fast turnover, the mice were allowed to survive 3 weeks after ending the BrdU delivery. Midbrain sections were processed for tyrosine hydroxylase (TH) immunohistochemistry, post-fixed with 4% paraformaldehyde, denaturated with 2N HCl and 0.025% pepsin, followed by immunolabelling of nuclear BrdU in single-stranded DNA. Double-labelled cells were analysed in a confocal laser microscope and showed segmented nuclear BrdU-label surrounded by TH-immunoreactive cytoplasm, never displaying apoptotic morphological features. Nigral neuronal proliferation was confirmed using another marker [3H]thymidine, delivered via an intraperitoneal osmotic pump. The protocol was well tolerated by the mice and not found to be toxic for the region studied, i.e. did not alter the total number of nigral neurons identified with TH/cresyl violet immunohistochemistry.

Robust Label Propagation on Multiple Networks

Transductive inference on graphs such as label propagation algorithms is receiving a lot of attention. In this paper, we address a label propagation problem on multiple networks and present a new algorithm that automatically integrates structure information brought in by multiple networks. The proposed method is robust in that irrelevant networks are automatically deemphasized, which is an advantage over Tsuda's approach (2005). We also show that the proposed algorithm can be interpreted as an expectation-maximization (EM) algorithm with a student-t prior. Finally, we demonstrate the usefulness of our method in protein function prediction and digit classification, and show analytically and experimentally that our algorithm is much more efficient than existing algorithms.

Copyright labeling of digitized image data

This article addresses the problem of embedding a robust copyright label into image data for identifying the ownership and distribution of the digitized image. The article first defines a reasonable set of functional requirements and design criteria for an image copyright labeling system, and then describes a labeling method using a JPEG model-based, frequency-hopped, randomly sequenced pulse position modulated code (RSPPMC). Our experiments demonstrated that the proposed method meets the requirements for copyright protection, including the robustness of an embedded label against several damaging possibilities such as lossy data compression, low-pass filtering, and color space conversion, as well as printing/scanning, scaling, and cropping.