Search

Post-secondary educational attainment is an increasingly important prerequisite to many valued outcomes. This paper examines the association of racially-specific high school course of study with student's postsecondary educational attainment using ELS 2002-2012 and a comprehensive measure of course intensity derived from students' patterns of course-taking. Results support a partial presence of racially-specific association of high school course of study. We also find that only Asian-American and American Indian/Alaskan Native students in the middle course intensity and Asian-American students in high course intensity ranges are more likely to obtain at least a 4-year college degree than same-race students in the low course intensity group. However, the same pattern is not evident for White, African-American, multi-racial, and Hispanic students, indicating that they do not experience the same benefit from academically rigorous courses of study. Also, educational expectations do not mediate the relationships among race, course of study, and postsecondary educational trajectory differently for students of different racial groups, i.e., they do not play a larger role in mediating the relationship between course of study and postsecondary educational trajectory for White students.

In recent years, researchers have focused on reducing the model size and number of computations (measured as “multiply-accumulate” or MAC operations) of DNNs. The energy consumption of a DNN depends on both the number of MAC operations and the energy efficiency of each MAC operation. The former can be estimated at design time; however, the latter depends on the intricate data reuse patterns and underlying hardware architecture. Hence, estimating it at design time is challenging. This article shows that the conventional approach to estimate the data reuse, viz. arithmetic intensity, does not always correctly estimate the degree of data reuse in DNNs since it gives equal importance to all the data types. We propose a novel model, termed “data type aware weighted arithmetic intensity” (DI), which accounts for the unequal importance of different data types in DNNs. We evaluate our model on 25 state-of-the-art DNNs on two GPUs. We show that our model accurately models data-reuse for all possible data reuse patterns for different types of convolution and different types of layers. We show that our model is a better indicator of the energy efficiency of DNNs. We also show its generality using the central limit theorem. [ABSTRACT FROM AUTHOR]

The remarkable predictive performance of deep neural networks (DNNs) has led to their adoption in service domains of unprecedented scale and scope. However, the widespread adoption and growing commercialization of DNNs have underscored the importance of intellectual property (IP) protection. Devising techniques to ensure IP protection has become necessary due to the increasing trend of outsourcing the DNN computations on the untrusted accelerators in cloud-based services. The design methodologies and hyper-parameters of DNNs are crucial information, and leaking them may cause massive economic loss to the organization. Furthermore, the knowledge of DNN's architecture can increase the success probability of an adversarial attack where an adversary perturbs the inputs and alters the prediction. In this work, we devise a two-stage attack methodology "DeepPeep," which exploits the distinctive characteristics of design methodologies to reverse-engineer the architecture of building blocks in compact DNNs. We show the efficacy of "DeepPeep" on P100 and P4000 GPUs. Additionally, we propose intelligent design maneuvering strategies for thwarting IP theft through the DeepPeep attack and proposed "SecureMobileNetV1." Interestingly, compared to vanilla MobileNet-V1, secure MobileNet-V1 provides a significant reduction in inference latency (≈60%) and improvement in predictive performance (≈2%) with very low memory and computation overheads. [ABSTRACT FROM AUTHOR]