Search

To defend against various adversarial attacks, it is essential to develop a robust and high computing efficiency defence framework. Adversarial ensemble learning is the most effective technique for defending against adversarial example attacks, which constructs ensembles of multiple Deep neural networks (DNNs) with adversarial training to obtain stronger defense. However, ensemble models run noticeably slower on existing DNN accelerators than single-model inference. Deploying ensemble models on the existing DNN accelerators leads to many critical issues such as the underutilization of hardware resources. To tackle emerging challenges, we propose EnsGuard, a dynamic asymmetric multicore systolic array architecture for adversarial ensemble learning inference to fully exploit both static and dynamic parallelism of ensemble models. Specifically, on the hardware level, we propose a novel instruction set extension and develop efficient architecture components to fully exploit the new hardware abstraction of scattered idle computing cores, and use them to dynamically create on-the-fly neural processing units (fNPUs). Moreover, we propose a computing power recycle mechanism to run on-the-fly models (small models) on fNPUs by carefully orchestrating execution order of ensemble models for maximizing hardware resources and bandwidth utilization. On the software level, EnsGuard adopts an integrated hardware/randomized ensemble co-design optimizer aiming at winning both faster inference and higher adversarial robustness. On top of that, a multimodel mapping method based on decision tree is proposed to enable the interleaving of different DNN executions both spatially and temporally, and mitigate straggler problems. Evaluation with a diverse set of workloads shows significant gains in throughput ( $4.4\times $ ) and energy reduction ( $3.2\times $ ).

Campylobacterjejuniis an important foodborne pathogen worldwide; however, there is a lack of information on the prevalence and antibiotic-resistant profile of C. jejuniin the People's Republic of China. We determined the prevalence and characteristics of C. jejunion the retail level in Tianjin, one of the five national central cities in China. A total of 227 samples of chicken wings, legs, and breasts were collected from supermarkets and wet markets; 42 of these samples were confirmed to be positive for Campylobactercontamination. The contamination rates of C. jejuniand other Campylobacterspecies were 13.7% (31 of 227 samples) and 5.7% (13 of 227 samples), respectively. A group of 31 C. jejuniisolates was subjected to antimicrobial susceptibility testing. All (100%) the selected isolates were resistant to ciprofloxacin and nalidixic acid; 77.4% were resistant to tetracycline, 67.7% to doxycycline, 35.5% to gentamicin, 25.8% to clindamycin and florfenicol, 19.4% to chloramphenicol, and 12.9% to erythromycin and azithromycin. A remarkably high proportion (41.9%) of multidrug-resistant isolates was identified. Multilocus sequence typing was conducted to study the population structure of the C. jejunistrains and their relationship to human isolates. The correlation between antimicrobial resistance traits and certain sequence types (STs) or clonal complexes was determined as well. A great genetic diversity of poultry isolates was identified, with 11 STs belonging to 6 clonal complexes and 11 singleton STs. The novel STs accounted for 40.9% (n= 9) of the 22 STs. ST-21, ST-353, ST-354, ST-443, ST-607, and ST-828 complexes had been previously identified from human isolates. This study revealed an extensive level of antimicrobial resistance and genetic diversity in C. jejuniisolated from chicken products in Tianjin, highlighting the necessity of performing enforced interventions to reduce Campylobacterprevalence in China.

AbstractThe aim of the study was to investigate the molecular epidemiologic of Staphylococcus aureusisolated from seven staphylococcal food-poisoning (SFP) outbreaks between 2006 and 2013 in Xi'an, northwest China. A total of seven S. aureusisolates associated with seven SFP outbreaks were obtained and characterized by determining the multilocus sequence typing, spatyping, pulsed-field gel electrophoresis (PFGE), antimicrobial susceptibility, toxin and resistant genes. The results showed that S. aureusST6-t701 (71.4%) was the most predominant clone, followed by ST5-t002 and ST59-t172 (each 14.3%). Seven different PFGE patterns were obtained. Moreover, resistance was most frequently observed to trimethoprim and penicillin (each 71.4%), followed by erythromycin (28.6%), ampicillin, clindamycin, and tetracycline (each 14.3%). All strains were susceptible to chloramphenicol, cefoxitin, oxacillin, amikacin, and vancomycin. Three of seven strains displayed resistance to three or more antimicrobials. Resistance genes were found as follows: linA/linA′(100%), blaZ(85.7%), tet(K), ermC, ermT, and ermB(each 14.3%). Other resistant genes were not detected. In addition, the most frequently identified exotoxin genes were seu, lukED, hla, hlb, hld(each 100%), followed by hlgand hlgv(each 85.7%), lukPV(71.4%), sea(57.1%), see(42.9%), etd(28.6%), and seb, sec, sed, sej, and sek(each 14.3%). The results indicated that S. aureusST6-t701, with a high prevalence in the northwest of China, exhibited multidrug resistance and harbored multiple toxin and resistance genes. Therefore, strict hygienic and preventative measures should be taken in order to avoid the contamination of food by S. aureusand toxin production in foods.