Explore the vast range of titles available.

Motion segmentation and non-rigid structure from motion are two challenging computer vision problems that have attracted numerous research interests. While the previous works handle these two problems separately, we present a general motion segmentation framework in this paper for solving these two seemingly different problems in a unified manner. At the heart of our general motion segmentation framework is a model selection mechanism based on finding the minimal basis subspace representation, by seeking the joint sparse representation of the data matrix. However, such formulation is NP-hard and we solve the convex proxy instead. Unlike other compressive sensing related works, this convex proxy solution is insufficient for our problem. The convex relaxation artefacts and noise yield multiple subspace representations, making identification of the exact number of motion subspaces challenging. We solve for the right number of subspaces by transforming this problem into a Facility Location problem with global cost and solve the factor graph formulation using max product belief propagation message passing.

In training deep learning networks, the optimizer and related learning rate are often used without much thought or with minimal tuning, even though it is crucial in ensuring a fast convergence to a good quality minimum of the loss function that can also generalize well on the test dataset. Drawing inspiration from the successful application of cyclical learning rate policy to computer vision tasks, we explore how cyclical learning rate can be applied to train transformer-based neural networks for neural machine translation. From our carefully designed experiments, we show that the choice of optimizers and the associated cyclical learning rate policy can have a significant impact on the performance. In addition, we establish guidelines when applying cyclical learning rates to neural machine translation tasks.

Low-rank matrix completion is a problem of immense practical importance. Recent works on the subject often use nuclear norm as a convex surrogate of the rank function. Despite its solid theoretical foundation, the convex version of the problem often fails to work satisfactorily in real-life applications. Real data often suffer from very few observations, with support not meeting the randomness requirements, ubiquitous presence of noise and potentially gross corruptions, sometimes with these simultaneously occurring. This paper proposes a Proximal Alternating Robust Subspace Minimization method to tackle the three problems. The proximal alternating scheme explicitly exploits the rank constraint on the completed matrix and uses the ℓ 0 pseudo-norm directly in the corruption recovery step. We show that the proposed method for the non-convex and non-smooth model converges to a stationary point. Although it is not guaranteed to find the global optimal solution, in practice we find that our algorithm can typically arrive at a good local minimizer when it is supplied with a reasonably good starting point based on convex optimization. Extensive experiments with challenging synthetic and real data demonstrate that our algorithm succeeds in a much larger range of practical problems where convex optimization fails, and it also outperforms various state-of-the-art algorithms.

Low-rank matrix completion is a problem of immense practical importance. Recent works on the subject often use nuclear norm as a convex surrogate of the rank function. Despite its solid theoretical foundation, the convex version of the problem often fails to work satisfactorily in real-life applications. Real data often suffer from very few observations, with support not meeting the random requirements, ubiquitous presence of noise and potentially gross corruptions, sometimes with these simultaneously occurring. This paper proposes a Proximal Alternating Robust Subspace Minimization (PARSuMi) method to tackle the three problems. The proximal alternating scheme explicitly exploits the rank constraint on the completed matrix and uses the \\(\\ell_0\\) pseudo-norm directly in the corruption recovery step. We show that the proposed method for the non-convex and non-smooth model converges to a stationary point. Although it is not guaranteed to find the global optimal solution, in practice we find that our algorithm can typically arrive at a good local minimizer when it is supplied with a reasonably good starting point based on convex optimization. Extensive experiments with challenging synthetic and real data demonstrate that our algorithm succeeds in a much larger range of practical problems where convex optimization fails, and it also outperforms various state-of-the-art algorithms.

In training deep learning networks, the optimizer and related learning rate are often used without much thought or with minimal tuning, even though it is crucial in ensuring a fast convergence to a good quality minimum of the loss function that can also generalize well on the test dataset. Drawing inspiration from the successful application of cyclical learning rate policy for computer vision related convolutional networks and datasets, we explore how cyclical learning rate can be applied to train transformer-based neural networks for neural machine translation. From our carefully designed experiments, we show that the choice of optimizers and the associated cyclical learning rate policy can have a significant impact on the performance. In addition, we establish guidelines when applying cyclical learning rates to neural machine translation tasks. Thus with our work, we hope to raise awareness of the importance of selecting the right optimizers and the accompanying learning rate policy, at the same time, encourage further research into easy-to-use learning rate policies.

Amaranthus tricolor L., a vegetable Amaranthus species, is an economically important crop containing large amounts of betalains. Betalains are natural antioxidants and can be classified into betacyanins and betaxanthins, with red and yellow colors, respectively. A. tricolor cultivars with varying betalain contents, leading to striking red to green coloration, have been commercially produced. However, the molecular differences underlying betalain biosynthesis in various cultivars of A. tricolor remain largely unknown. In this study, A. tricolor cultivars with different colors were chosen for comparative transcriptome analysis. The elevated expression of AmCYP76AD1 in a red-leaf cultivar of A. tricolor was proposed to play a key role in producing red betalain pigments. The functions of AmCYP76AD1 , AmDODAα1 , AmDODAα2 , and AmcDOPA5GT were also characterized through the heterologous engineering of betalain pigments in Nicotiana benthamiana . Moreover, high and low L-DOPA 4,5-dioxygenase activities of AmDODAα1 and AmDODAα2, respectively, were confirmed through in vitro enzymatic assays. Thus, comparative transcriptome analysis combined with functional and enzymatic studies allowed the construction of a core betalain biosynthesis pathway of A. tricolor . These results not only provide novel insights into betalain biosynthesis and evolution in A. tricolor but also provide a basal framework for examining genes related to betalain biosynthesis among different species of Amaranthaceae .

Background Informal caregiving is an integral part of post-stroke recovery with strenuous caregiving demands often resulting in caregiving burden, threatening sustainability of caregiving and potentially impacting stroke survivor’s outcomes. Our study aimed to examine and quantify objective and subjective informal care burden after stroke; and to explore the factors associated with informal care burden in Singapore. Methods Stroke patients and their informal caregivers were recruited from all five tertiary hospitals in Singapore from December 2010 to September 2013. Informal care comprised of assistance provided by informal caregivers with any of the activities of daily living. Informal care burden was measured by patients’ likelihood of requiring informal care, hours of informal care required, and informal caregivers’ Zarit’s Burden Score. We examined informal care burden at 3-months and 12-months post-stroke. Generalized linear regressions were applied with control variables including patients’ and informal caregivers’ demographic characteristics, arrangement of informal care, and patients’ health status including stroke severity (measured using National Institute of Health Stroke Scale), functional status (measured using Modified Rankin Scale), self-reported depression, and common comorbidities. Results Three hundred and five patients and 263 patients were examined at 3-months and 12-months. Around 35% were female and 60% were Chinese. Sixty three percent and 49% of the patients required informal care at 3-months and 12-months point, respectively. Among those who required informal care, average hours required per week were 64.3 h at 3-months and 76.6 h at 12-months point. Patients with higher functional dependency were more likely to require informal care at both time points, and required more hours of informal care at 3-months point. Female informal caregivers and those caring for patients with higher functional dependency reported higher Zarit’s Burden. While informal caregivers who worked full-time reported higher burden, those caring for married stroke patients reported lower burden at 3-months point. Informal caregivers who co-cared with foreign domestic workers, i.e.: stay-in migrant female waged domestic workers, reported lower burden. Conclusions Informal care burden remains high up to 12-months post-stroke. Factors such as functional dependency, stroke severity, informal caregiver gender and co-caring with foreign domestic workers were associated with informal care burden.

Investigations showed a haemoglobin concentration of 8·1 g/dL, platelet count of 565 × 109 per L (normal 150–400), and white cell count of 10·5 × 109 per L; the mean corpuscular volume was 77·2 fL (normal 80–94) and the mean corpuscular haemoglobin was 23·1 pg per cell—suggestive of iron deficiency anaemia. Immunohistochemistry stains demonstrated strong and diffuse nuclear immunoreactivity to NKX2-2 and ERG, with membranous immunoreactivity to CD99, and an absence of nuclear immunoreactivity to FLI1: supporting a possible fusion of the EWSRI gene on chromosome 22q12 with the gene encoding the transcription factor ERG (appendix). Young adults presenting with recurrent vomiting, weight loss, an abdominal mass, and anaemia should raise the suspicion of an aggressive malignancy (video).

This study aimed to quantify health-related quality of life (HRQoL) loss associated with first episode of stroke by comparing patient-reported HRQoL before and after stroke onset. The impact of stroke in local population was also evaluated by comparing the pre- and post-stroke HRQoL with that of the general population. The HRQoL of stroke survivors was assessed with the EQ-5D-3L index score at recruitment, for recalled pre-stroke HRQoL, and at 3 and 12 month post-stroke. Change in HRQoL from pre-stroke to 3 and 12 month was self-reported by 285 and 238 patients, respectively. Mean EQ index score at each time point (baseline: 464 patients; 3 month post-stroke: 306 patients; 12 month post-stroke: 258 patients) was compared with published population norms for EQ-5D-3L. There was a significant decrease in HRQoL at 3 (0.25) and 12 month (0.09) post-stroke when compared to the retrospectively recalled patients' mean pre-stroke HRQoL level (0.87). The reduction at 3 month was associated with the reduction in all EQ-5D-3L health dimensions; reductions remaining at 12 month were limited to dimensions of mobility, self-care, usual activities, and anxiety/depression. Stroke patients had a lower mean EQ index than the general population by 0.07 points pre-stroke (0.87 vs. 0.94), 0.33 points at 3 month (0.61 vs. 0.94) and 0.18 points at 12 month (0.76 vs. 0.94) post-stroke. Stroke has a substantial impact on HRQoL in Singapore, especially in the first three months post-stroke. Compared to the general population, stroke survivors have lower HRQoL even before stroke onset. This pre-stroke deficit in HRQoL should be taken into account when quantifying health burden of stroke or setting goals for stroke rehabilitation.