Implementation of energy-efficient convolutional neural networks based on kernel-pruned silicon photonics.

Silicon-based optical neural networks offer the prospect of high-performance computing on integrated photonic circuits. However, the scalability of on-chip optical depth networks is restricted by the limited energy and space resources. Here, we present a silicon-based photonic convolutional neural network (PCNN) combined with the kernel pruning, in which the optical convolutional computing core of PCNN is a tunable micro-ring weight bank. Our numerical simulation demonstrates the effect of weight mapping accuracy on PCNN performance and we find that the performance of PCNN decreases significantly when the weight mapping accuracy is less than 4.3 bits. Additionally, the experimental demonstration shows that the accuracy of the PCNN on the MNIST dataset has a slight loss compared to the original CNN when 93.75 % of the convolutional kernels are pruned. By making use of kernel pruning, the energy saved by a convolutional kernel removal is about 202.3 mW, and the overall energy saved has a linear relationship with the number of kernels removed. The methodology is scalable and provides a feasible solution for implementing faster and more energy-efficient large-scale optical convolutional neural networks on photonic integrated circuits.

Validation of Herek's attitudes toward lesbian women and gay men scale among undergraduates in mainland China.

The lack of a standardized reliable and valid instrument makes it difficult to measure attitudes toward lesbian women and gay men (ATLG) consistently and thus poses a challenge to compare and contrast intervention measures. This study aimed to validate Herek's ATLG scale among undergraduates in mainland China and identify factors associated with negative attitudes toward LG. A total of 6,036 eligible undergraduates conveniently drawn from 30 provinces across mainland China were randomly split in half. Item analysis was first used to select unrelated or redundant items for deletion. Exploratory factor analysis (EFA) were then conducted on the first half of the sample (n = 3,001), followed by confirmatory factor analysis (CFA) and reliability analysis in the second half (n = 3035). Logistic regression analyses were finally carried out to identify their determinants. Six items were removed from the item analysis. EFA supported the existence of two factors (ATL and ATG). CFA results indicated that the two-factor model fit the data better than the one-factor model. Logistic regression analyses indicated that being female, majoring in non-health-related disciplines, attributing homosexuality to uncontrollable causes, non-adherence to traditional gender norms and exposure to homosexual content were significantly associated with less negative attitudes toward both L and G. Urban students were marginally less likely to express negative attitudes toward L but not G, while non-heterosexuals and those who had prior personal contact with homosexuals exhibited less negative attitudes toward G but not L. However, grade showed no significant associations with either ATL or ATG. The retained 14-item version of Herek's ATLG scale has been proven to be a reliable and valid tool. Furthermore, ATL and ATG were determined by different factors and thus would be treated separately. In order to reduce negative attitudes toward LG among undergraduates in mainland China, a comprehensive intervention plan such as conducting comprehensive sex education and pushing the process of legalizing same-sex marriage should be designed, implemented and evaluated.

Discordance between perceived risk and actual risky sexual behaviors among undergraduate university students in mainland China: a cross-sectional study.

BACKGROUND: HIV prevention, diagnosis, treatment and care services might be hampered by inaccurate risk assessment. This study aimed to investigate the extent of and factors associated with the discordance between perceived risk and actual risky sexual behaviors among undergraduates in mainland China, guided by the Anderson's behavioral model. METHODS: This study involved a secondary analysis of cross-sectional data collected during the fall semester of 2018-2019 academic year. The present analysis was restricted to 8808 undergraduates with low risk perception. Those who had low perceived risk but actually engaged in risky sexual behaviors were categorized as risk discordance (RD). Univariate and multivariate Logistic regression analyses were conducted to identify factors associated with RD. RESULTS: Overall, the discordance rate between perceived and actual risk was 8.5% (95% CI: 7.9%-9.1%). Multivariate Logistic regression analysis indicated that non-heterosexual women (AOR = 0.41, 95% CI:0.27-0.60), heterosexual men (AOR = 0.45, 95% CI:0.33-0.61) and women (AOR = 0.26, 95% CI:0.19-0.35) were less likely to exhibit RD, when compared with non- heterosexual men. Furthermore, non-freshmen (AOR = 1.57, 95% CI:1.30-1.90), early initiators of sexual intercourse (AOR = 5.82, 95% CI:4.10-8.26), and those who had lower levels of HIV knowledge (AOR = 1.28, 95% CI:1.08-1.51), displayed higher levels of stigma against PLHIV (AOR = 1.50, 95% CI:1.26-1.77) and had ever been tested for HIV (AOR = 1.36, 95% CI:1.04-1.77) were more prone to reporting RD. Those with more enabling resources [i.e., displaying high levels of condom use self-efficacy (AOR = 0.70, 95% CI:0.59-0.84) and being knowledge of local testing center (AOR = 0.71, 95% CI:0.60-0.83)] were less likely to report RD. However, spending more than 2000 Yuan a month on basic needs (AOR = 2.55, 95% CI:2.07-3.14), residing in urban areas (AOR = 1.35, 95% CI:1.15-1.59) and being knowledgeable of the national AIDS policy (AOR = 1.40,95% CI:1.18-1.66) increased the chance of exhibiting RD. CONCLUSIONS: Comprehensive interventions, including targeting students with high-risk characteristics, improving the acceptability of PrEP and PEP, conducting health education, enhancing self-efficacy for using condoms and making opt-out HIV testing routine in college campus, should be taken to reduce the discordance between perceived and actual HIV risk and finally to reach the goal of Zero AIDS.

Inequities in the utilization of HIV counseling and testing services among undergraduates in mainland China.

BACKGROUND: HIV counseling and testing (HCT) is provided free of charge on college campuses, but very few studies have examined whether HCT uptake is equitably distributed. This cross-sectional study aimed to compare the relative importance of various determinants of HCT uptake among undergraduates in mainland China so as to assess and subsequently to suggest ways to eliminate inequities in its use, guided by the Andersen's behavioral model. METHODS: A total of 10,665 eligible undergraduates were conveniently selected to complete an online survey. Hierarchical logistic regression analyses were performed to identify the factors associated with HCT utilization. RESULTS: Only 7.7% of undergraduates had utilized HCT services. HCT uptake was inequitably distributed, since it was mainly determined by predisposing and enabling factors rather than by need factors. Further analysis indicated that HCT uptake was significantly related to two need factors, one enabling factor and six predisposing factors. Those with a higher need [i.e., perceiving themselves to be at higher risk of acquiring HIV infection (AOR = 2.76, 95% CI:2.02-3.78) and engaging in condomless sex (AOR = 1.29, 95% CI:1.00-1.67)] and those with more resources [i.e., being knowledgeable of local AIDS service organization (AOR = 1.59, 95% CI:1.37-1.85)] were more likely to utilize HCT services. Compared to non-heterosexual men, non-heterosexual women (AOR = 0.51, 95% CI:0.37-0.72), heterosexual men (AOR = 0.44, 95% CI:0.33-0.57) and women (AOR = 0.31, 95%CI: 0.24-0.41) were less likely to utilize HCT service. Furthermore, those with more knowledge (AOR = 0.80, 95% CI:0.69-0.94) and taking a positive attitude towards HCT services [i.e, expressing their willingness to utilize HCT services (AOR = 0.68, 95% CI:0.56-0.81) and having recognized the necessity to provide HTC services in the local university (AOR = 0.46, 95% CI:0.36-0.57)] were less likely to utilize HCT services. However, medical students (AOR = 1.34, 95% CI: 1.15-1.56) and non-freshmen (AOR = 1.22, 95% CI:1.03-1.45) were more likely to utilize HCT services. CONCLUSIONS: To increase HCT uptake and simultaneously reduce the remaining inequities, a comprehensive intervention should be continued to target non-heterosexual men and non- freshmen and those with a higher need for HCT services, conduct health education, improve the availability and accessibility of HIV testing services.

Application of the Chinese version of Zelaya's HIV-related stigma scale to undergraduates in mainland China.

BACKGROUND: This cross-sectional study aims to validate the Chinese version of Zelaya's HIV-related Stigma Scale (CVZHSS) among a large undergraduate sample in mainland China, and apply it to measure the level of different dimensions of stigma and their respective determinants. METHODS: From September 10, 2018, to January 9, 2019, a total of 10,665 eligible undergraduates conveniently drawn from 30 provinces in mainland China (except for Tibet) completed the self-designed online questionnaire distributed via sojump.com voluntarily, anonymously and confidentially. Both exploratory and confirmatory factor analyses (EFA and CFA) were first performed to test its construct validity, Cronbach's alpha was then used to assess its internal consistency, and Logistic regression analyses were finally carried out to identify predictors of various dimensions of stigma. RESULTS: As expected from the original model, four factors (i.e., "fear of casual transmission", "moral judgment", "personal stigma" and "perceived community stigma") were extracted using principal component analysis with varimax rotation, accounting for 63.26% of the total variance. The CFA further confirmed the four-factor construct (CFI = 0.92, GFI = 0.91, RMSEA = 0.07). In addition, all the four factors demonstrated acceptable internal consistency with Cronbach's alpha ranging from 0.83 to 0.92. Stigma as measured by "fear of casual transmission" (74.4%), "moral judgement" (61.6%), "personal stigma" (79.0%) and "perceived community stigma"(36.5%) is highly prevalent among undergraduates. Except for non-freshmen, less knowledge about HIV and unsafe sex which were consistently associated with higher levels of stigma in all four dimensions, other eight variables including gender, residential area, major, sexual orientation, having ever being tested perception of HIV risk, willingness to utilize HTC service and awareness of the national AIDS policy played differential roles in affecting different dimensions of stigma. CONCLUSIONS: The CVZHSS is a reliable and valid measurement tool and can be used to identify undergraduates with high levels of stigma. However, the four dimensions (Fear, moral judgement, personal stigma and perceived community stigma) were respectively influenced by different determinants, and thus should be treated independently when designing, implementing and evaluating stigma reduction programs.