University of Adelaide

About: University of Adelaide is a education organization based out in Adelaide, South Australia, Australia. It is known for research contribution in the topics: Population & Poison control. The organization has 27251 authors who have published 79167 publications receiving 2671128 citations. The organization is also known as: The University of Adelaide & Adelaide University.

Methodological quality of case series studies: an introduction to the JBI critical appraisal tool.

Abstract: Introduction Systematic reviews provide a rigorous synthesis of the best available evidence regarding a certain question. Where high-quality evidence is lacking, systematic reviewers may choose to rely on case series studies to provide information in relation to their question. However, to date there has been limited guidance on how to incorporate case series studies within systematic reviews assessing the effectiveness of an intervention, particularly with reference to assessing the methodological quality or risk of bias of these studies. Methods An international working group was formed to review the methodological literature regarding case series as a form of evidence for inclusion in systematic reviews. The group then developed a critical appraisal tool based on the epidemiological literature relating to bias within these studies. This was then piloted, reviewed and approved by the international Scientific Committee of JBI. Results The JBI critical appraisal tool for case series studies includes 10 questions addressing the internal validity and risk of bias of case series designs, particularly confounding, selection and information bias, in addition to the importance of clear reporting. Conclusion In certain situations, case series designs may represent the best available evidence to inform clinical practice. The JBI critical appraisal tool for case series offers systematic reviewers an approved method to assess the methodological quality of these studies.

High concentrations of Na+ and Cl– ions in soil solution have simultaneous detrimental effects on growth of faba bean under salinity stress

Abstract: Despite the fact that most plants accumulate both sodium (Na(+)) and chloride (Cl(-)) ions to high concentration in their shoot tissues when grown in saline soils, most research on salt tolerance in annual plants has focused on the toxic effects of Na(+) accumulation. There have also been some recent concerns about the ability of hydroponic systems to predict the responses of plants to salinity in soil. To address these two issues, an experiment was conducted to compare the responses to Na(+) and to Cl(-) separately in comparison with the response to NaCl in a soil-based system using two varieties of faba bean (Vicia faba), that differed in salinity tolerance. The variety Nura is a salt-sensitive variety that accumulates Na(+) and Cl(-) to high concentrations while the line 1487/7 is salt tolerant which accumulates lower concentrations of Na(+) and Cl(-). Soils were prepared which were treated with Na(+) or Cl(-) by using a combination of different Na(+) salts and Cl(-) salts, respectively, or with NaCl. While this method produced Na(+)-dominant and Cl(-)-dominant soils, it unavoidably led to changes in the availability of other anions and cations, but tissue analysis of the plants did not indicate any nutritional deficiencies or toxicities other than those targeted by the salt treatments. The growth, water use, ionic composition, photosynthesis, and chlorophyll fluorescence were measured. Both high Na(+) and high Cl(-) reduced growth of faba bean but plants were more sensitive to Cl(-) than to Na(+). The reductions in growth and photosynthesis were greater under NaCl stress and the effect was mainly additive. An important difference to previous hydroponic studies was that increasing the concentrations of NaCl in the soil increased the concentration of Cl(-) more than the concentration of Na(+). The data showed that salinity caused by high concentrations of NaCl can reduce growth by the accumulation of high concentrations of both Na(+) and Cl(-) simultaneously, but the effects of the two ions may differ. High Cl(-) concentration reduces the photosynthetic capacity and quantum yield due to chlorophyll degradation which may result from a structural impact of high Cl(-) concentration on PSII. High Na(+) interferes with K(+) and Ca(2+) nutrition and disturbs efficient stomatal regulation which results in a depression of photosynthesis and growth. These results suggest that the importance of Cl(-) toxicity as a cause of reductions in growth and yield under salinity stress may have been underestimated.

Structured Knowledge Distillation for Semantic Segmentation

Abstract: In this paper, we investigate the issue of knowledge distillation for training compact semantic segmentation networks by making use of cumbersome networks. We start from the straightforward scheme, pixel-wise distillation, which applies the distillation scheme originally introduced for image classification and performs knowledge distillation for each pixel separately. We further propose to distill the structured knowledge from cumbersome networks into compact networks, which is motivated by the fact that semantic segmentation is a structured prediction problem. We study two such structured distillation schemes: (i) pair-wise distillation that distills the pairwise similarities, and (ii) holistic distillation that uses adversarial training to distill holistic knowledge. The effectiveness of our knowledge distillation approaches is demonstrated by extensive experiments on three scene parsing datasets: Cityscapes, Camvid and ADE20K.

STRIP: a defence against trojan attacks on deep neural networks

Abstract: A recent trojan attack on deep neural network (DNN) models is one insidious variant of data poisoning attacks. Trojan attacks exploit an effective backdoor created in a DNN model by leveraging the difficulty in interpretability of the learned model to misclassify any inputs signed with the attacker's chosen trojan trigger. Since the trojan trigger is a secret guarded and exploited by the attacker, detecting such trojan inputs is a challenge, especially at run-time when models are in active operation. This work builds STRong Intentional Perturbation (STRIP) based run-time trojan attack detection system and focuses on vision system. We intentionally perturb the incoming input, for instance by superimposing various image patterns, and observe the randomness of predicted classes for perturbed inputs from a given deployed model---malicious or benign. A low entropy in predicted classes violates the input-dependence property of a benign model and implies the presence of a malicious input---a characteristic of a trojaned input. The high efficacy of our method is validated through case studies on three popular and contrasting datasets: MNIST, CIFAR10 and GTSRB. We achieve an overall false acceptance rate (FAR) of less than 1%, given a preset false rejection rate (FRR) of 1%, for different types of triggers. Using CIFAR10 and GTSRB, we have empirically achieved result of 0% for both FRR and FAR. We have also evaluated STRIP robustness against a number of trojan attack variants and adaptive attacks.