We address the challenges associated with deploying neural networks on CPUs, with a particular focus on minimizing inference time while maintaining accuracy. Our novel approach is to use the dataflow ...(i.e., computation order) of a neural network to explore data reuse opportunities using heuristic-guided analysis and a code generation framework, which enables exploration of various Single Instruction, Multiple Data (SIMD) implementations to achieve optimized neural network execution. Our results demonstrate that the dataflow that keeps outputs in SIMD registers while also maximizing both input and weight reuse consistently yields the best performance for a wide variety of inference workloads, achieving up to 3x speedup for 8-bit neural networks, and up to 4.8x speedup for binary neural networks, respectively, over the optimized implementations of neural networks today.
Recent quantization techniques have enabled heterogeneous precisions at very fine granularity, e.g., each parameter/activation can take on a different precision, resulting in compact neural networks ...without sacrificing accuracy. However, there is a lack of efficient architectural support for such networks, which require additional hardware to decode the precision settings for individual variables, align the variables, and provide fine-grained mixed-precision compute capabilities. The complexity of these operations introduces high overheads. Thus, the improvements in inference latency/energy of these networks are not commensurate with the compression ratio, and may be inferior to larger quantized networks with uniform precisions. We present an end-to-end co-design approach encompassing computer architecture, training algorithm, and inference optimization to efficiently execute networks with fine-grained heterogeneous precisions. The key to our approach is a novel training algorithm designed to accommodate hardware constraints and inference operation requirements, outputting networks with input-channel-wise heterogeneous precisions and at most three precision levels. Combined with inference optimization techniques, existing architectures with low-cost enhancements can support such networks efficiently, yielding optimized tradeoffs between accuracy, compression ratio and inference latency/energy. We demonstrate the efficacy of our approach across CPU and GPU architectures. For various representative neural networks, our approach achieves >10x improvements in both compression ratio and inference latency, with negligible degradation in accuracy compared to full-precision networks.
Crowdsourcing is a computing paradigm where humans are actively involved in a computing task, especially for tasks that are intrinsically easier for humans than for computers. Spatial crowdsourcing ...is an increasing popular category of crowdsourcing in the era of mobile Internet and sharing economy, where tasks are spatiotemporal and must be completed at a specific location and time. In fact, spatial crowdsourcing has stimulated a series of recent industrial successes including sharing economy for urban services (Uber and Gigwalk) and spatiotemporal data collection (OpenStreetMap and Waze). This survey dives deep into the challenges and techniques brought by the unique characteristics of spatial crowdsourcing. Particularly, we identify four core algorithmic issues in spatial crowdsourcing: (1) task assignment, (2) quality control, (3) incentive mechanism design, and (4) privacy protection. We conduct a comprehensive and systematic review of existing research on the aforementioned four issues. We also analyze representative spatial crowdsourcing applications and explain how they are enabled by these four technical issues. Finally, we discuss open questions that need to be addressed for future spatial crowdsourcing research and applications.
Full text
Available for:
EMUNI, FIS, FZAB, GEOZS, GIS, IJS, IMTLJ, KILJ, KISLJ, MFDPS, NLZOH, NUK, OBVAL, OILJ, PNG, SAZU, SBCE, SBJE, SBMB, SBNM, UKNU, UL, UM, UPUK, VKSCE, ZAGLJ
First-principles calculations of e− ph interactions are becoming a pillar of electronic structure theory. However, the current approach is incomplete. The piezoelectric (PE) e− ph interaction, a ...long-range scattering mechanism due to acoustic phonons in noncentrosymmetric polar materials, is not accurately described at present. Current calculations include short-range e− ph interactions (obtained by interpolation) and the dipolelike Frölich long-range coupling in polar materials, but lack important quadrupole effects for acoustic modes and PE materials. Here we derive and compute the long-range e− ph interaction due to dynamical quadrupoles, and apply this framework to investigate e− ph interactions and the carrier mobility in the PE material wurtzite GaN. We show that the quadrupole contribution is essential to obtain accurate e− ph matrix elements for acoustic modes and to compute PE scattering. Our work resolves the outstanding problem of correctly computing e− ph interactions for acoustic modes from first principles, and enables studies of e− ph coupling and charge transport in PE materials.
Full text
Available for:
CMK, CTK, FMFMET, IJS, NUK, PNG, UL, UM
The well accepted "free drug hypothesis" for small-molecule drugs assumes that only the free (unbound) drug concentration at the therapeutic target can elicit a pharmacologic effect. Unbound (free) ...drug concentrations in plasma are readily measurable and are often used as surrogates for the drug concentrations at the site of pharmacologic action in pharmacokinetic-pharmacodynamic analysis and clinical dose projection in drug discovery. Furthermore, for permeable compounds at pharmacokinetic steady state, the free drug concentration in tissue is likely a close approximation of that in plasma; however, several factors can create and maintain disequilibrium between the free drug concentration in plasma and tissue, leading to free drug concentration asymmetry. These factors include drug uptake and extrusion mechanisms involving the uptake and efflux drug transporters, intracellular biotransformation of prodrugs, membrane receptor-mediated uptake of antibody-drug conjugates, pH gradients, unique distribution properties (covalent binders, nanoparticles), and local drug delivery (e.g., inhalation). The impact of these factors on the free drug concentrations in tissues can be represented by
, the ratio of free drug concentration between tissue and plasma at steady state. This review focuses on situations in which free drug concentrations in tissues may differ from those in plasma (e.g.,
> or <1) and discusses the limitations of the surrogate approach of using plasma-free drug concentration to predict free drug concentrations in tissue. This is an important consideration for novel therapeutic modalities since systemic exposure as a driver of pharmacologic effects may provide limited value in guiding compound optimization, selection, and advancement. Ultimately, a deeper understanding of the relationship between free drug concentrations in plasma and tissues is needed.
The broad impact of the COVID-19 on self-reported daily behaviors and health in Chinese and US samples remains unknown. This study aimed to compare physical and mental health between people from the ...United States (U.S.) and China, and to correlate mental health parameters with variables relating to physical symptoms, knowledge about COVID-19, and precautionary health behaviors. To minimize risk of exposure, respondents were electronically invited by existing study respondents or by data sourcing software and surveys were completed via online survey platforms. Information was collected on demographics, physical symptoms, contact history, knowledge about COVID-19, psychologic parameters (i.e. IES-R; DASS-21), and health behaviors. The study included a total of 1445 respondents (584 U.S.; 861 China). Overall, Americans reported more physical symptoms, contact history, and perceived likelihood of contracting COVID-19. Americans reported more stress and depressive symptoms, while Chinese reported higher acute-traumatic stress symptoms. Differences were identified regarding face mask use and desires for COVID-19 related health information, with differential mental health implications. Physical symptoms that were possibly COVID-19 related were associated with adverse mental health. Overall, American and Chinese participants reported different mental and physical health parameters, health behaviors, precautionary measures, and knowledge of COVID-19; different risk and protective factors were also identified.
Full text
Available for:
DOBA, EMUNI, FIS, FZAB, GEOZS, GIS, IJS, IMTLJ, IZUM, KILJ, KISLJ, MFDPS, NLZOH, NUK, OILJ, PILJ, PNG, SAZU, SBCE, SBJE, SBMB, SBNM, UILJ, UKNU, UL, UM, UPUK, VKSCE, VSZLJ, ZAGLJ
The novel Coronavirus-2019 (COVID-19) was declared a pandemic by the World Health Organization (WHO) in March 2020, impacting the lifestyles, economy, physical and mental health of individuals ...globally. This study aimed to test the model triggered by physical symptoms resembling COVID-19 infection, in which the need for health information and perceived impact of the pandemic mediated the path sequentially, leading to adverse mental health outcomes. A cross-sectional research design with chain mediation model involving 4612 participants from participating 8 countries selected via a respondent-driven sampling strategy was used. Participants completed online questionnaires on physical symptoms, the need for health information, the Impact of Event Scale-Revised (IES-R) questionnaire and Depression, Anxiety and Stress Scale (DASS-21). The results showed that Poland and the Philippines were the two countries with the highest levels of anxiety, depression and stress; conversely, Vietnam had the lowest mean scores in these areas. Chain mediation model showed the need for health information, and the perceived impact of the pandemic were sequential mediators between physical symptoms resembling COVID-19 infection (predictor) and consequent mental health status (outcome). Excessive and contradictory health information might increase the perceived impact of the pandemic. Rapid COVID-19 testing should be implemented to minimize the psychological burden associated with physical symptoms, whilst public mental health interventions could target adverse mental outcomes associated with the pandemic.
Full text
Available for:
IZUM, KILJ, NUK, PILJ, PNG, SAZU, UL, UM, UPUK