Backpropagation-Based Decoding for Multimodal Machine Translation

People are able to describe images using thousands of languages, but languages share only one visual world. The aim of this work is to use the learned intermediate visual representations from a deep convolutional neural network to transfer information across languages for which paired data is not available in any form. Our work proposes using backpropagation-based decoding coupled with transformer-based multilingual-multimodal language models in order to obtain translations between any languages used during training. We particularly show the capabilities of this approach in the translation of German-Japanese and Japanese-German sentence pairs, given a training data of images freely associated with text in English, German, and Japanese but for which no single image contains annotations in both Japanese and German. Moreover, we demonstrate that our approach is also generally useful in the multilingual image captioning task when sentences in a second language are available at test time. The results of our method also compare favorably in the Multi30k dataset against recently proposed methods that are also aiming to leverage images as an intermediate source of translations.

Dealing with spatial heterogeneity in pointwise-to-gridded- data comparisons

Most studies on validation of satellite trace gas retrievals or atmospheric chemical transport models assume that pointwise measurements, which roughly represent the element of space, should compare well with satellite (model) pixels (grid box). This assumption implies that the field of interest must possess a high degree of spatial homogeneity within the pixels (grid box), which may not hold true for species with short atmospheric lifetimes or in the proximity of plumes. Results of this assumption often lead to a perception of a nonphysical discrepancy between data, resulting from different spatial scales, potentially making the comparisons prone to overinterpretation. Semivariogram is a mathematical expression of spatial variability in discrete data. Modeling the semivariogram behavior permits carrying out spatial optimal linear prediction of a random process field using kriging. Kriging can extract the spatial information (variance) pertaining to a specific scale, which in turn translates pointwise data to a gridded space with quantified uncertainty such that a grid-to-grid comparison can be made. Here, using both theoretical and real-world experiments, we demonstrate that this classical geostatistical approach can be well adapted to solving problems in evaluating model-predicted or satellite-derived atmospheric trace gases. This study suggests that satellite validation procedures using the present method must take kriging variance and satellite spatial response functions into account. We present the comparison of Ozone Monitoring Instrument (OMI) tropospheric NO2 columns against 11 Pandora spectrometer instrument (PSI) systems during the DISCOVER-AQ campaign over Houston. The least-squares fit to the paired data shows a low slope (OMI=0.76×PSI+1.18×1015 molecules cm−2, r2=0.66), which is indicative of varying biases in OMI. This perceived slope, induced by the problem of spatial scale, disappears in the comparison of the convolved kriged PSI and OMI (0.96×PSI+0.66×1015 molecules cm−2, r2=0.72), illustrating that OMI possibly has a constant systematic bias over the area. To avoid gross errors in comparisons made between gridded data vs. pointwise measurements, we argue that the concept of semivariogram (or spatial autocorrelation) should be taken into consideration, particularly if the field exhibits a strong degree of spatial heterogeneity at the scale of satellite and/or model footprints.

Home-based unsupervised pulmonary rehabilitation program improves the respiratory disability in systemic sclerosis patients with dyspnea: an observational prospective study

Dyspnea is a common symptom in Systemic Sclerosis (SSc) that considerably decreases patients’ quality of life (QoL). Pulmonary Rehabilitation (PR) mitigates dyspnea impact on daily activities. The aim of this study is to evaluate the effect on respiratory disability of home-based PR in SSc patients with dyspnea. In this observational prospective monocentric study, we screened all dyspneic SSc consecutive patients attending the Rheumatological day hospital in the University hospital of Parma from January 2019 and June 2019. The aim of our study was to understand if a PR unsupervised home-based program could improve respiratory disability in this specific population. Dyspnea was evaluated with the self-administered questionnaires modified Medical Research Council (mMRC) and Saint George's Respiratory Questionnaire (SGRQ).Patients also filled in Short Form 36 (SF36) and the Modified-Health Assessment Questionnaire for SSc (HAQ-MOD). Health Professionals assessed and trained the patients and collected data before PR and at the end of the program. PR consisted in 5 weekly unsupervised sessions for 8 weeks. Wilcoxon test for paired data evaluated the changes after PR. p<0.05 was considered statistically significant. 46 SSc patients were included (43 female). Only 31 (29 female) performed PR as planned (Adherent Group-AG) while the others gave up within the first week (Non-Adherent Group-NAG). All SGRQ domains (Symptoms: from 30 to 18; p=0.0055; Activity: from 47 to 35, p=0.23; Impact from 29 to 25, p=0.044) and SGRQ total score (from 35 to 29; p=0.022) improved in AG. SGRQ scores did not change in NAG as well as SF36 and HAQ-MOD in both groups. The home-based PR program dramatically decreased the effect, frequency and severity of respiratory symptoms. Conversely, it slightly changed the activities causing breathlessness and dyspnea-related social functioning disturbances. PR appears to be a useful tool in treatment strategies aiming to achieve a QoL improvement in SSc patients.

Unpaired Underwater Image Enhancement Based on CycleGAN

Underwater image enhancement recovers degraded underwater images to produce corresponding clear images. Image enhancement methods based on deep learning usually use paired data to train the model, while such paired data, e.g., the degraded images and the corresponding clear images, are difficult to capture simultaneously in the underwater environment. In addition, how to retain the detailed information well in the enhanced image is another critical problem. To solve such issues, we propose a novel unpaired underwater image enhancement method via a cycle generative adversarial network (UW-CycleGAN) to recover the degraded underwater images. Our proposed UW-CycleGAN model includes three main modules: (1) A content loss regularizer is adopted into the generator in CycleGAN, which constrains the detailed information existing in one degraded image to remain in the corresponding generated clear image; (2) A blur-promoting adversarial loss regularizer is introduced into the discriminator to reduce the blur and noise in the generated clear images; (3) We add the DenseNet block to the generator to retain more information of each feature map in the training stage. Finally, experimental results on two unpaired underwater image datasets produced satisfactory performance compared to the state-of-the-art image enhancement methods, which proves the effectiveness of the proposed model.

Cellular and transcriptional responses of SH-SY5Y human neurocytes following in vitro exposure to Gelsemium sempervirens

Background: Gelsemium sempervirens (Gelsemium s.) is a highly toxic plant but is employed at low doses and/or high dilutions as an anxiolytic and antidepressant. Previous investigations in our laboratory [1,2] have shown a significant anxiolytic-like activity of Gelsemium s., using emotional response models in laboratory mice. Although there is some biochemical evidence of a possible role of neurosteroid metabolism [3], the cellular and molecular mechanisms involved in the effects of Gelsemium s. at the level of nervous system are largely unknown. To help determine these pathways, we used human neurocytes (SH-SY5Y cell line) treated in vitro with different dilutions of Gelsemium s. and evaluated their vitality and gene expression changes. Methods: The drugs were produced by Boiron Laboratoires, Lyon (F), starting from a whole-plant-hydroalcoholic extract of Gelsemium s. Solutions 1C, 2C, 3C, 4C, 8C and 29C (C= centesimal dilution/dynamization prepared in 30% ethanol/distilled water) were provided in 30-ml glass bottles, wrapped in aluminium foil and were stored in the dark at room temperature in a metal cupboard. Control solutions (“placebo”) were serially diluted/dynamized 30% ethanol/distilled water. Before each experiment, 0.05 ml samples of Gelsemium s. and placebo were added to 5 ml of distilled sterile and apyrogenic water in a 15 ml Falcon polystyrene plastic tube, closed and shaken in mechanical shaker DinaA for 7.5 sec (150 strokes) to obtain the final 2C, 3C, 4C, 5C, 9C and 30C succussed dilutions, with ethanol concentration of 0.3% (v/v) (final 0.03% in the assay system). Human neuroblastoma cell line SHSY5Y was grown in DMEM-F12 medium (Lonza), with 10% fetal bovine serum (FBS), penicillin (100 units/ml) and streptomycin (100 mg/ml). To assess cell viability and metabolism, 20,000 cells per well were seeded in 96 microplate wells in 200 μl of medium. After overnight incubation, 22 μl of drug or placebo were added and the plate was incubated at 37°C with 5% CO2 in a humidified atmosphere for 72 hours. Then the viability test with reagent WST-1 (Roche) was performed for 3 hours and the absorbance was detected with multiplate reader. A total of 17 experiments, each done with six replicate microwells. To make a relative measurement of protein, the cells were lysed and a Bradford assay was done directly in the plate. The Student t-test and the sign rank test for paired data were utilized for data analysis. To obtain a profile of gene expression, cells were pre preconditioned with Gelsemium s./placebo dilutions for 24 h, then RNA was isolated and analysed by microarray and RT-PCR. SHSY5Y cells were plated onto Petri dishes (day 1) and the day after the medium was replaced with the medium with 2% FBS (day 2). After 24h, 10%v/v Gelsemium s. or placebo dilutions were added to the medium (day 3) and cells were incubated for a further 24h. On day 4, cells were then harvested and the RNA extracted using the Qiagen RNAeasy Mini Kit following the manufacturer’s instructions. Microarray analysis was performed on a custom 12 x 135 k human NimbleGen microarray containing 45033 genes with 3 probes per target gene. Four biological replicates were analysed for each condition. Analysis of differentially expressed genes was performed using linear modelling and empirical Bayes methods and p-values were adjusted for multiple testing with the Benjamini and Hochberg method. A Human Neurotransmitter Receptors and Regulators RT2 Profiler PCR array (Qiagen) was performed in profiling the expression of genes involved in modulating the biological processes of neurotransmitter biosynthesis, uptake, transport and signaling through neurotransmitter receptors. Results: In viability tests, cells treated with Gelsemium s. showed slightly higher metabolic activity (3-4 %) than those treated with placebo. Overall comparison of the data for the whole sample of placebo versus that of Gelsemium s. using the Student t-test showed a small but significant difference (p < 0.001). Furthermore, a non parametric approach comparing the two treatments at the same dilution yielded a significant difference under the sign test (p < 0.01) and Wilcoxon rank test for paired data (p < 0.05), so that the values of differences were also considered. The differences between groups having the same dilution (placebo 2C versus Gelsemium s. 2C etc.) were significant in four dilutions: 2 C, 3 C, 4C (p < 0.01), 9 C (p < 0.02), while 5 C and 30 C yielded non-significant values. No changes due to Gelsemium s. were detected using protein assay, suggesting that the viability test revealed effects on metabolic activity instead of on cell proliferation. In microarray analysis, transcripts expression was analyzed and genes differentially expressed by the Gelsemium s. dilutions were selected. A gene was considered to be differentially expressed if it showed an absolute value of log-ratio greater than or equal to 0.5, an index that translates to a fold-change of 1.4 in transcript quantity. Out of a total of 45,033 transcripts, exposure to Gelsemium s. 2C promoted the selective downexpression of 49 genes (p values adj

Interrupting the crossover effect of leader's role overload on employee's negative affect: the moderating role of self-concordance goal

PurposeThe purpose of the study is to examine the crossover effect of leader's role overload on employee's negative affect. More importantly, the stuy will identify the buffering role of self-concordance goal on the relationship between leader's role overload and employee's negative affect.Design/methodology/approachThe study builds the crossover impact of leader's role overload on employee's negative affect as well as the moderating effect of self-concordance goal. By a two-wave and paired data from 51 leaders and 225 employees, the study examines the hypothesis using cross-level analysis.FindingsResults show that leader's role overload tends to reduce negative affect for employees who pursue high-level self-concordance goal and increase negative affect for employees who pursue low-level self-concordance goal.Practical implicationsIt is important for employees to get rid of negative affect in the workplace. The study informs managers the benefits of pursuing self-concordance goals in helping employees alleviate the negative effect of leader's role overload.Originality/valueFindings of the present study can enrich the literature of the crossover process from leader to employee and offer management strategy for enterprises about how to buffer the damaging effect of leader's role overload on employees.

Social media usage and employee creativity: is relational energy a missing link?

PurposePrevious studies overemphasize the negative effects of social media usage (SMU) within organizations and underestimate its positive influences on employees' behavior. This study attempts to link employees' social media use at work to their creativity performance.Design/methodology/approachBased on the bounded generalized reciprocity theory and unbounded indirect reciprocity (UIR) theory, the authors developed a research model. To test the model, the authors collected a set of 172 paired data of organizations and employees from 31 knowledge-intensive enterprises in China to test the hypothesis.FindingsThis research found that the social, cognitive and hedonic uses of social media all directly affect employee creativity. Relational energy fully mediates the effects of the cognitive and hedonic usages on creativity. Moreover, job autonomy moderates the effects of the relationships among the social, cognitive and hedonic uses on employee creativity.Originality/valueThe conclusions not only enriched authors’ understanding of the effectiveness of interpersonal interaction but also extended the research boundary of the relationship between SMU and employee creativity.

Probing the effect on student conceptual understanding due to a forced mid-semester transition to online teaching

The Force Concept Inventory (FCI) can be used as an assessment tool to measure conceptual gains in a cohort of students. The FCI uses a conceptions/``misconceptions" lens rather than a context dependent perspective, such as ``knowledge-in-pieces". In this study it was given to first year students ($N=256$ students) pre- and post-mechanics lectures, at the University of Johannesburg. From these results we examine the effect of switching mid-semester from traditional classes to online classes, as imposed by the COVID-19 lockdown in South Africa. Overall results indicate no appreciable difference of gain, when bench-marked against previous studies using this assessment tool. When compared with $2019$ grades, the $2020$ semester grades do not appear to be greatly affected. Furthermore, statistical analyses also indicate a gender difference in mean gains in favour of females at the $95\%$ significance level (for paired data, $N=48$).

The Regularity of the Site of Impaction in Recurrent Gallstone Ileus: A Systematic Review and Meta-Analysis of Reported Cases

Objective. Due to the rarity of recurrent gallstone ileus (RGSI), its epidemiological and clinical features are elusive. With a focus on mortality and the site of impaction, this study consolidates the key clinical characteristics of index GSI (IGSI) and RGSI. Methods. A meta-analysis of cases reported on RGSI was performed. Risk factors for mortality and site of impaction were examined, and a subgroup analysis was performed for age, sex, and site of impaction (jejunum, ileum, or others). Results. In the final analysis, 50 (56 individual cases) studies were included. The paired data for the site of impaction was available for 45 patients. Women accounted for 87.3% of all RGSI cases included in the pooled analysis. The median age (interquartile range, IQR) of the patients was 70 (63–76) years, and the median time of recurrence (IQR) was 20.5 (8.5–95.5) days. The overall mortality rate was 11.8%, without correlation between the mortality rate and age, the time of recurrence, or the site of impaction. The region in which the stone was found in RGSI and IGSI was similar in most cases p = 0.002 . Logistic regression also revealed a higher probability of stone impaction in the ileum in RGSI if it was the site of impaction in IGSI. In most cases, enterolithotomy was the preferred method. Conclusions. A high index of suspicion for RGSI should be maintained for older women with a history of GSI. The region where the stone was impacted during IGSI should be investigated first in such patients.