Effects of age and flight experience on prefrontal cortex activity in airline pilots: An fNIRS study.

It is essential for airlines to have a deep understanding of the cognitive impact of aging among pilots. The current literature on executive function indicates that compensatory mechanisms in the brain may counteract age-related cognitive decline, at least up to certain task load levels. However, few studies have been administered to evaluate changes in aircrew competence as they age. The present study focuses on dorsolateral prefrontal cortex (DLPFC) activity as it is implicated in cognitive performance and working memory, which are associated with skill proficiency. We measured the DLPFC activity for airline pilots, including trainees, during maneuvering using a flight simulator. Our preliminary results indicated that only expert (aged) pilots demonstrated higher activity of the left DLPFC than the right one. However, for youth trainees, not only was the error rate high while using the flight simulator, but the activity of the DLFPC was also lower than that of the expert pilots, and there was no statistically significant difference between the left and right DLPFC. Although these findings partially differ from those reported in previous studies on age-related changes, it is evident that training as an airline pilot for over 20 years may affect such results. We believe that this noninvasive approach to objective quantification of skill will facilitate the development of effective assessment competence in aging.

Tackling the curse of dimensionality with physics-informed neural networks.

The curse-of-dimensionality taxes computational resources heavily with exponentially increasing computational cost as the dimension increases. This poses great challenges in solving high-dimensional partial differential equations (PDEs), as Richard E. Bellman first pointed out over 60 years ago. While there has been some recent success in solving numerical PDEs in high dimensions, such computations are prohibitively expensive, and true scaling of general nonlinear PDEs to high dimensions has never been achieved. We develop a new method of scaling up physics-informed neural networks (PINNs) to solve arbitrary high-dimensional PDEs. The new method, called Stochastic Dimension Gradient Descent (SDGD), decomposes a gradient of PDEs' and PINNs' residual into pieces corresponding to different dimensions and randomly samples a subset of these dimensional pieces in each iteration of training PINNs. We prove theoretically the convergence and other desired properties of the proposed method. We demonstrate in various diverse tests that the proposed method can solve many notoriously hard high-dimensional PDEs, including the Hamilton-Jacobi-Bellman (HJB) and the Schrödinger equations in tens of thousands of dimensions very fast on a single GPU using the PINNs mesh-free approach. Notably, we solve nonlinear PDEs with nontrivial, anisotropic, and inseparable solutions in less than one hour for 1000 dimensions and in 12 h for 100,000 dimensions on a single GPU using SDGD with PINNs. Since SDGD is a general training methodology of PINNs, it can be applied to any current and future variants of PINNs to scale them up for arbitrary high-dimensional PDEs.

Predominance of T-bet-positive Th1 cells in infiltrating T-lymphocytes in most of active arteritis lesions of giant cell arteritis.

Immunological mechanisms through the activation of CD4-positive T-cells have been assumed to be involved in the pathogenesis of giant cell arteritis (GCA). Many studies employing frozen tissues of temporal artery biopsy, peripheral blood lymphocytes, and plasma of GCA patients have revealed the contribution of interferon-γ and interleukin-17 in both protein and mRNA levels. However, the analyses using formalin-fixed and paraffin-embedded (FFPE) tissue specimens, in which the correlation between histopathologic pictures and immunological circumstances would be elucidated, have been limited. Here, we performed the immunohistochemical analyses of infiltrating small lymphocytes in GCA lesions using FFPE specimens, especially of the subsets of CD4-positive T-cells by immunohistochemistry with antibodies against T-bet, GATA-3, RORγT, and Foxp3, which is the differentiation-specific transcription factor for Th1, Th2, Th17, and Treg cells, respectively. In these slides, the nuclear-positive staining is much more clearly and easily identifiable than the cytoplasmic staining for cytokines. The results indicate the predominance of T-bet-positive Th1 cells in infiltrating T-cells in most of active arteritis lesions of GCA. Furthermore, our data suggest the possible immunosuppressive microenvironment induced by T-reg cells and M2-type macrophages in the arteritis lesions throughout the course of GCA inflammation.

Time-course analysis of pulmonary inflammation induced by intratracheal instillation of nanosized crystalline silica particles in F344 rats.

Crystalline silica is an important cause of serious pulmonary diseases, and its toxic potential is known to be associated with its surface electrical properties. However, in vivo data clarifying the relevance of silica's toxic potential, especially its long-term effects, remain insufficient. To investigate the contribution of physico-chemical property including surface potential on the hazard of nanocrystalline silica, we performed single intratracheal instillation testing using five different crystalline silicas in a rat model and assessed time-course changes in pulmonary inflammation, lung burden, and thoracic lymph node loads. Silica-nanoparticles were prepared from two commercial products (Min-U-Sil5 [MS5] and SIO07PB [SPB]) using three different pretreatments: centrifugation (C), grinding (G), and surface dissolving (D). The five types of silica particles-MS5, MS5_C, SPB_C, SPB_G, and SPB_D-were intratracheally instilled into male F344 rats at doses of 0 mg/kg (purified water), 0.22 mg/kg (SPB), and 0.67, 2, or 6 mg/kg (MS5). Bronchoalveolar lavage, a lung burden analysis, and histopathological examination were performed at 3, 28, and 91 days after instillation. Granuloma formation was present in MS5 group at 91 days after instillation, although granuloma formation was suppressed in MS5_C group, which had a smaller particle size. SPB_C induced severe and progressive inflammation and kinetic lung overload, whereas SPB_G and SPB_D induced only slight and transient acute inflammation. Our results support that in vivo toxic potential of nanosilica by intratracheal instillation may involve with surface electrical properties leading to prolonged effect and may not be dependent not only on surface properties but also on other physico-chemical properties.

Interpolated Adversarial Training: Achieving robust neural networks without sacrificing too much accuracy.

Adversarial robustness has become a central goal in deep learning, both in the theory and the practice. However, successful methods to improve the adversarial robustness (such as adversarial training) greatly hurt generalization performance on the unperturbed data. This could have a major impact on how the adversarial robustness affects real world systems (i.e. many may opt to forego robustness if it can improve accuracy on the unperturbed data). We propose Interpolated Adversarial Training, which employs recently proposed interpolation based training methods in the framework of adversarial training. On CIFAR-10, adversarial training increases the standard test error ( when there is no adversary) from 4.43% to 12.32%, whereas with our Interpolated adversarial training we retain the adversarial robustness while achieving a standard test error of only 6.45%. With our technique, the relative increase in the standard error for the robust model is reduced from 178.1% to just 45.5%. Moreover, we provide mathematical analysis of Interpolated Adversarial Training to confirm its efficiencies and demonstrate its advantages in terms of robustness and generalization.

Understanding Dynamics of Nonlinear Representation Learning and Its Application.

Representations of the world environment play a crucial role in artificial intelligence. It is often inefficient to conduct reasoning and inference directly in the space of raw sensory representations, such as pixel values of images. Representation learning allows us to automatically discover suitable representations from raw sensory data. For example, given raw sensory data, a deep neural network learns nonlinear representations at its hidden layers, which are subsequently used for classification (or regression) at its output layer. This happens implicitly during training through minimizing a supervised or unsupervised loss. In this letter, we study the dynamics of such implicit nonlinear representation learning. We identify a pair of a new assumption and a novel condition, called the on-model structure assumption and the data architecture alignment condition. Under the on-model structure assumption, the data architecture alignment condition is shown to be sufficient for the global convergence and necessary for global optimality. Moreover, our theory explains how and when increasing network size does and does not improve the training behaviors in the practical regime. Our results provide practical guidance for designing a model structure; for example, the on-model structure assumption can be used as a justification for using a particular model structure instead of others. As an application, we then derive a new training framework, which satisfies the data architecture alignment condition without assuming it by automatically modifying any given training algorithm dependent on data and architecture. Given a standard training algorithm, the framework running its modified version is empirically shown to maintain competitive (practical) test performances while providing global convergence guarantees for deep residual neural networks with convolutions, skip connections, and batch normalization with standard benchmark data sets, including MNIST, CIFAR-10, CIFAR-100, Semeion, KMNIST, and SVHN.

[An autopsy case of nasal-type primary gastric extranodal NK/T-cell lymphoma with early liver failure].

An 82-year-old woman was admitted to the hospital because of tiredness and fever. She was diagnosed with acute hepatitis. Although the cause of acute hepatitis was undetermined, her health condition and liver function improved, and she was discharged. Four weeks later, she was hospitalized again because of anorexia. Laboratory data revealed worsened anemia. Endoscopy results revealed a huge ulcerative lesion in the lesser curvature of the stomach. After 4 days, she vomited blood and died of hemorrhagic shock. The autopsy revealed a nasal-type primary gastric extranodal NK/T-cell lymphoma (ENKTL). Although no lymphoma cells were found in the liver biopsy collected during the first hospitalization, lymphoma cells and lymphocytes in the liver tissue were identified during autopsy because the lymphoma had infiltrated the liver. Primary gastric ENKTL is extremely rare and poorly understood. However, the general prognosis of progressive ENKTL is poor. Early diagnosis of liver metastasis of lymphoma cells is difficult;thus, in some cases, lymphoma metastases to the liver are diagnosed during autopsy. Although further experiments are required, we report a rare case of primary gastric ENKTL.

Interpolation consistency training for semi-supervised learning.

We introduce Interpolation Consistency Training (ICT), a simple and computation efficient algorithm for training Deep Neural Networks in the semi-supervised learning paradigm. ICT encourages the prediction at an interpolation of unlabeled points to be consistent with the interpolation of the predictions at those points. In classification problems, ICT moves the decision boundary to low-density regions of the data distribution. Our experiments show that ICT achieves state-of-the-art performance when applied to standard neural network architectures on the CIFAR-10 and SVHN benchmark datasets. Our theoretical analysis shows that ICT corresponds to a certain type of data-adaptive regularization with unlabeled points which reduces overfitting to labeled points under high confidence values.

Two Nicotiana occidentalis accessions enable gene identification for Type II hybrid lethality by the cross to N. sylvestris.

Hybrid lethality, meaning the death of F1 hybrid seedlings, has been observed in many plant species, including Nicotiana. Previously, we have revealed that hybrids of the selected Nicotiana occidentalis accession and N. tabacum, an allotetraploid with S and T genomes, exhibited lethality characterized by the fading of shoot color. The lethality was suggested to be controlled by alleles of loci on the S and T genomes derived from N. sylvestris and N. tomentosiformis, respectively. Here, we extended the analysis of hybrid lethality using other two accessions of N. occidentalis identified from the five tested accessions. The two accessions were crossed with N. tabacum and its two progenitors, N. sylvestris and N. tomentosiformis. After crosses with N. tabacum, the two N. occidentalis accessions yielded inviable hybrid seedlings whose lethality was characterized by the fading of shoot color, but only the T genome of N. tabacum was responsible for hybrid lethality. Genetic analysis indicated that first-mentioned N. occidentalis accession carries a single gene causing hybrid lethality by allelic interaction with the S genome.

Studies on analogs of DAMASCENOLIDETM: Part 4. Synthesis and odor evaluation of sulfur-containing analogs of DAMASCENOLIDETM.

DAMASCENOLIDETM [1, 4-(4-methylpent-3-en-1-yl)furan-2(5H)-one], which is isolated from damask rose, is a useful aroma compound with a citrus-like odor. We have previously reported on the synthesis and odor properties of 34 analogs of 1 as part of our new aroma compound development project. In order to develop better aroma compounds and to gather more information on structure-odor relationships, 6 novel sulfur-containing analogs of 1 were synthesized. Odor evaluation revealed that their odors differed significantly from those of the corresponding sulfur-free compounds. The introduction of a sulfur atom does not necessarily result in a sulfur-like odor. In particular, the 2(5H)-thiophenone analogs gave waxy, oily, and lactone-like odors that are uncharacteristic of sulfur-containing compounds. In many synthesized analogs, the introduction of a sulfur atom led to an increase in odor intensity, as expected.

Time-course comparison of pulmonary inflammation induced by intratracheal instillation of four different nickel oxide nanoparticles in male Fischer rats.

Occupational exposure to nickel oxide (NiO) is an important cause of respiratory tract cancer. Toxicity is known to be associated with the dissociated component, i.e. nickel (II) ions. To address the relationship between physicochemical properties, including solubility in artificial lysosomal fluid, of NiO and time-course changes in the pulmonary response, we conducted an intratracheal instillation study in male Fischer rats using four different well-characterized NiO products, US3352 (NiO A), NovaWireNi01 (NiO B), I small particle (NiO C), and 637130 (NiO D). The NiOs were suspended in purified water and instilled once intratracheally into male F344 rats (12 weeks old) at 0 (vehicle control), 0.67, 2, and 6 mg/kg body weight. The animals were euthanized on days 3, 28, or 91 after instillation, and blood analysis, bronchoalveolar lavage fluid (BALF) testing, and histopathological examination were performed. The most soluble product, NiO B, caused the most severe systemic toxicity, leading to a high mortality rate, but the response was transient and surviving animals recovered. The second-most-soluble material, NiO D, and the third, NiO A, caused evident pulmonary inflammation, and the responses persisted for at least 91 days with collagen proliferation. In contrast, NiO C induced barely detectable inflammation in the BALF examination, and no marked changes were noted on histopathology. These results indicate that the early phase toxic potential of NiO products, but not the persistence of pulmonary inflammation, is associated with their solubility.

Studies on analogs of DAMASCENOLIDETM: Part 3. Synthesis and odor evaluation of dimethylated, cyclopropanated, and other analogs of DAMASCENOLIDETM.

DAMASCENOLIDETM [1, 4-(4-methylpent-3-en-1-yl)furan-2(5H)-one], which has a citrus-like odor, is an important aroma component of roses. We have previously reported on the synthesis and odor evaluation of 24 analogs of 1 as part of our new aroma compound developing project. To accumulate more information on structure-odor relationships, 10 more promising analogs such as dimethylated and cyclopropanated analogs were synthesized and subjected to odor evaluation. As a result, it was found that dimethylation of the furanone ring affected the odor. It was also found that cyclopropanation of 1 affected the odor, whereas cyclopropanation of the double bond isomer of 1 did not significantly affect the odor. The effects on the odor caused by ring size expansion and replacement of the side chain were also investigated.

Comprehensive gene expression profiling of human astrocytes treated with a hepatic encephalopathy-inducible factor, alpha 1-antichymotripsin.

Astrocytes are major glial cells that play a critical role in brain homeostasis. Abnormalities in astrocytic function, such as hepatic encephalopathy (HE) during acute liver failure, can result in brain death following brain edema and the associated astrocyte swelling. Recently, we have identified alpha 1-antichymotripsin (ACT) to be a biomarker candidate for HE. ACT induces astrocyte swelling by upregulating aquaporin 4 (AQP4); however, the causal connection between these proteins is not clear yet. In this study, we utilized a microarray profile to screen the differentially expressed genes (DEGs) in astrocytes treated with ACT. We then performed Gene Ontology, REACTOME, and the comprehensive resource of mammalian protein complexes (CORUM) enrichment analyses of the identified DEGs. The results of these analyses indicated that the DEGs were enriched in pathways activating adenylate cyclase (AC)-coupled G protein-coupled receptors (GPCRs) and therefore were involved in the cyclic adenosine monophosphate (cAMP) signaling. These results indicate that ACT may act as a ligand of Gs-GPCRs and subsequently upregulate cAMP. As cAMP is known to upregulate AQP4 in astrocytes, these results suggest that ACT may upregulate AQP4 by activating AC GPCRs and therefore serve as a therapeutic target for acute HE.

Locally adaptive activation functions with slope recovery for deep and physics-informed neural networks.

We propose two approaches of locally adaptive activation functions namely, layer-wise and neuron-wise locally adaptive activation functions, which improve the performance of deep and physics-informed neural networks. The local adaptation of activation function is achieved by introducing a scalable parameter in each layer (layer-wise) and for every neuron (neuron-wise) separately, and then optimizing it using a variant of stochastic gradient descent algorithm. In order to further increase the training speed, an activation slope-based slope recovery term is added in the loss function, which further accelerates convergence, thereby reducing the training cost. On the theoretical side, we prove that in the proposed method, the gradient descent algorithms are not attracted to sub-optimal critical points or local minima under practical conditions on the initialization and learning rate, and that the gradient dynamics of the proposed method is not achievable by base methods with any (adaptive) learning rates. We further show that the adaptive activation methods accelerate the convergence by implicitly multiplying conditioning matrices to the gradient of the base method without any explicit computation of the conditioning matrix and the matrix-vector product. The different adaptive activation functions are shown to induce different implicit conditioning matrices. Furthermore, the proposed methods with the slope recovery are shown to accelerate the training process.

Immunostaining With Immunoglobulin G Subclass Antibody Cocktail for Diagnosis of Type 1 Autoimmune Pancreatitis.

BACKGROUND.: Immunoglobulin (Ig) G4-related diseases (RDs) are systemic diseases in which serum IgG4 levels are frequently elevated. They can cause diffuse or focal tumor formation, organ swelling, and tissue thickening in organs infiltrated by IgG4+ plasma cells. The diagnostic criteria for IgG4-RDs include an IgG4/IgG ratio >40%, but counting IgG+ cells can be difficult because of the weakness of IgG staining density. We hypothesized that an antibody cocktail of mixed IgG1, IgG2, IgG3, and IgG4 (AC-IgG) might give immunohistochemistry results comparable with those of IgG in IgG4-RD. METHODS.: We compared AC-IgG reactivity with IgG expression in type 1 autoimmune pancreatitis (AIP), a representative IgG4-RD. We compared immunohistochemistry results using AC-IgG and IgG-only in 10 cases of AIP. The coefficient of variation (Cv) was used to analyze differences between AC-IgG and IgG findings in AIP by 13 board-certified pathologists. RESULTS.: Although mean values for IgG+ cells did not significantly differ between AC-IgG (34.3; range = 27.4-37.1) and IgG (30.0; range = 23.0-45.6; P = .6254), Cv was lower for AC-IgG (33.4%) than for IgG (51.4%; regression equation; y[IgG] = 0.988x + 0.982; correlation coefficient = 0.907). The data showed that the results of both methods were largely consistent. CONCLUSION.: AC-IgG could replace IgG to count IgG+ cells because of its lower Cv.

Analog synthesis of DAMASCENOLIDETM, an important aroma component of roses, and their odor properties.

DAMASCENOLIDETM [1, 4-(4-methylpent-3-en-1-yl)furan-2(5H)-one], which has a citrus-like odor, is an important aroma component of roses. We have previously reported on the synthesis and odor evaluation of double-bond isomers of 1 and concluded that the position and the geometric isomerism of the double-bond had a significant effect on the odor. For the purpose of deepening knowledge about structure-odor relationships, we synthesized 13 analogs of compound 1 and evaluated their odors. As a result, it was found that the presence of two double-bonds and branched methyl group at the terminal position in the side chain was essential in order to have a citrus-like odor. Substitution of the side chain with appropriate length at the appropriate 4-position of the 2(5H)-furanone ring was also an important factor in determining the quality of the odor.

Tranexamic acid-induced fixed drug eruption confirmed by the drug lymphocyte transformation test.

We present the first case of multiple fixed drug eruption caused by tranexamic acid, which was confirmed by the LTT. This case was difficult to diagnose because the drug-induced aseptic meningitis by loxoprofen was occurred simultaneously.

Every Local Minimum Value Is the Global Minimum Value of Induced Model in Nonconvex Machine Learning.

For nonconvex optimization in machine learning, this article proves that every local minimum achieves the globally optimal value of the perturbable gradient basis model at any differentiable point. As a result, nonconvex machine learning is theoretically as supported as convex machine learning with a handcrafted basis in terms of the loss at differentiable local minima, except in the case when a preference is given to the handcrafted basis over the perturbable gradient basis. The proofs of these results are derived under mild assumptions. Accordingly, the proven results are directly applicable to many machine learning models, including practical deep neural networks, without any modification of practical methods. Furthermore, as special cases of our general results, this article improves or complements several state-of-the-art theoretical results on deep neural networks, deep residual networks, and overparameterized deep neural networks with a unified proof technique and novel geometric insights. A special case of our results also contributes to the theoretical foundation of representation learning.

Depth with nonlinearity creates no bad local minima in ResNets.

In this paper, we prove that depth with nonlinearity creates no bad local minima in a type of arbitrarily deep ResNets with arbitrary nonlinear activation functions, in the sense that the values of all local minima are no worse than the global minimum value of corresponding classical machine-learning models, and are guaranteed to further improve via residual representations. As a result, this paper provides an affirmative answer to an open question stated in a paper in the conference on Neural Information Processing Systems 2018. This paper advances the optimization theory of deep learning only for ResNets and not for other network architectures.