The prediction accuracy has been the long-lasting and sole standard for comparing the performance of different image classification models, including the ImageNet competition. However, recent studies have highlighted the lack of robustness in well-trained deep neural networks to adversarial examples. Visually imperceptible perturbations to natural images can easily be crafted and mislead the image classifiers towards misclassification. To demystify the trade-offs between robustness and accuracy, in this paper we thoroughly benchmark 18 ImageNet models using multiple robustness metrics, including the distortion, success rate and transferability of adversarial examples between 306 pairs of models. Our extensive experimental results reveal several new insights:(1) linear scaling law -the empirical 2 and ∞ distortion metrics scale linearly with the logarithm of classification error; (2) model architecture is a more critical factor to robustness than model size, and the disclosed accuracy-robustness Pareto frontier can be used as an evaluation criterion for ImageNet model designers; (3) for a similar network architecture, increasing network depth slightly improves robustness in ∞ distortion; (4) there exist models (in VGG family) that exhibit high adversarial transferability, while most adversarial examples crafted from one model can only be transferred within the same family. Experiment code is publicly available at https://github.com/huanzhang12/Adversarial Survey.where f (x, t) is a loss function to measure the distance between the prediction of x and the target label t. In this work, we choose f (x, t) = max{max i =t [(Logit(x)) i − (Logit(x)) t ], −κ}

Graph neural networks (GNNs) which apply the deep neural networks to graph data have achieved significant performance for the task of semisupervised node classification. However, only few work has addressed the adversarial robustness of GNNs. In this paper, we first present a novel gradient-based attack method that facilitates the difficulty of tackling discrete graph data. When comparing to current adversarial attacks on GNNs, the results show that by only perturbing a small number of edge perturbations, including addition and deletion, our optimization-based attack can lead to a noticeable decrease in classification performance. Moreover, leveraging our gradientbased attack, we propose the first optimizationbased adversarial training for GNNs. Our method yields higher robustness against both different gradient based and greedy attack methods without sacrificing classification accuracy on original graph. Code is available at https://github.com/ KaidiXu/GCN_ADV_Train.

No abstract

The global demand for food could double in another 40 y owing to growth in the population and food consumption per capita. To meet the world's future food and sustainability needs for biofuels and renewable materials, the production of starch-rich cereals and cellulose-rich bioenergy plants must grow substantially while minimizing agriculture's environmental footprint and conserving biodiversity. Here we demonstrate one-pot enzymatic conversion of pretreated biomass to starch through a nonnatural synthetic enzymatic pathway composed of endoglucanase, cellobiohydrolyase, cellobiose phosphorylase, and alpha-glucan phosphorylase originating from bacterial, fungal, and plant sources. A special polypeptide cap in potato alpha-glucan phosphorylase was essential to push a partially hydrolyzed intermediate of cellulose forward to the synthesis of amylose. Up to 30% of the anhydroglucose units in cellulose were converted to starch; the remaining cellulose was hydrolyzed to glucose suitable for ethanol production by yeast in the same bioreactor. Next-generation biorefineries based on simultaneous enzymatic biotransformation and microbial fermentation could address the food, biofuels, and environment trilemma.bioeconomy | food and feed | synthetic amylose | in vitro synthetic biology | cell-free biomanufacturing T he continuing growth of the population and food consumption per capita means that the global demand for food could increase by 50-100% by 2050 (1, 2), and ∼30% of the world's agricultural land and 70% of the world's fresh water withdrawals are being used for the production of food and feed to support 7 billion people (3, 4). Starch is the most important dietary component because it accounts for more than half of the consumed carbohydrates, which provide 50-60% of the calories needed by humans. Starch is composed of polysaccharides consisting of a large number of glucose units joined together primarily by alpha-1,4-glycosidic bonds and alpha-1,6-glycosidic bonds. Linear-chain amylose is more valuable than branched amylopectin because it can be used as a precursor for making high-quality transparent, flexible, low-oxygen-diffusion plastic sheets and films (5, 6); tailored functional food or additives for lowering the risk of serious noninfectious diseases (e.g., diabetes and obesity) (7, 8); and a potential high-density hydrogen carrier (9-11). Also, it is easy to convert linear amylose to branched amylopectin by using alphaglucan-branching glycosyltransferase (12).Cellulose, a linear glucan linked by beta-1,4-glycosidic bonds, is the supporting material of plant cell walls and the most abundant carbohydrate on Earth. The annual resource of cellulosic materials is ∼40 times greater than the starch produced by crops cultivated for food and feed. In addition, (perennial) cellulosic plants and dedicated bioenergy crops can grow on low-quality land, even on marginal land, and require fewer inputs such as fertilizers, herbicides, pesticides, and water, whereas annual high-productivity starch-rich crops require high-qual...

No abstract