Joint Program and Layout Transformations to Enable Convolutional Operators on Specialized Hardware Based on Constraint Programming

The success of Deep Artificial Neural Networks (DNNs) in many domains created a rich body of research concerned with hardware accelerators for compute-intensive DNN operators. However, implementing such operators efficiently with complex hardware intrinsics such as matrix multiply is a task not yet automated gracefully. Solving this task often requires joint program and data layout transformations. First solutions to this problem have been proposed, such as TVM, UNIT, or ISAMIR, which work on a loop-level representation of operators and specify data layout and possible program transformations before the embedding into the operator is performed. This top-down approach creates a tension between exploration range and search space complexity, especially when also exploring data layout transformations such as im2col, channel packing, or padding. In this work, we propose a new approach to this problem. We created a bottom-up method that allows the joint transformation of both computation and data layout based on the found embedding. By formulating the embedding as a constraint satisfaction problem over the scalar dataflow, every possible embedding solution is contained in the search space. Adding additional constraints and optimization targets to the solver generates the subset of preferable solutions. An evaluation using the VTA hardware accelerator with the Baidu DeepBench inference benchmark shows that our approach can automatically generate code competitive to reference implementations. Further, we show that dynamically determining the data layout based on intrinsic and workload is beneficial for hardware utilization and performance. In cases where the reference implementation has low hardware utilization due to its fixed deployment strategy, we achieve a geomean speedup of up to × 2.813, while individual operators can improve as much as × 170.

Integrated Tick Management in Guilford, CT: Fipronil-Based Rodent-Targeted Bait Box Deployment Configuration and Peromyscus leucopus (Rodentia: Cricetidae) Abundance Drive Reduction in Tick Burdens

Integrated tick management (ITM) is a comprehensive strategy used to reduce presence of ticks and their associated pathogens. Such strategies typically employ a combination of host and non-host targeted treatments which often include fipronil-based, rodent-targeted bait boxes. Bait boxes target small-bodied rodents, specifically white-footed mice (Peromyscus leucopus Rafinesque) that not only play a crucial role in the blacklegged tick (Ixodes scapularis Say (Ixodida: Ixodidae)) life cycle, but also in the transmission of numerous pathogens, primarily Borrelia burgdorferi Johnson, Schmid, Hyde, Steigerwalt & Brenner (Spirochaetales: Spirochaetaceae), the causal agent of Lyme disease. This study aimed to determine the effect of bait box deployment configuration on tick burden reduction while also further exploring bait consumption and P. leucopus abundances as measures of bait box usage and effectiveness. Boxes were deployed on nine properties within each of six neighborhoods (n = 54) in two different configurations: grid and perimeter. Multiple factors were analyzed as potential predictors for reduction in tick burdens using a backward stepwise selection procedure. Results confirmed the perimeter configuration was a more effective deployment strategy. In addition, overall P. leucopus abundance was a significant predictor of tick burden reduction while bait consumption was not. These findings not only further support the recommended perimeter deployment configuration but provide insight into effective utilization in areas of high P. leucopus abundance. The identification of this significant relationship, in addition to configuration, can be utilized by vector control professionals and homeowners to make informed decisions on bait box placement to make sustained impacts on the I. scapularis vector and associated pathogens within an ITM framework.

From pandemic response to portable population health: A formative evaluation of the Detroit mobile health unit program

This article describes our experience developing a novel mobile health unit (MHU) program in the Detroit, Michigan, metropolitan area. Our main objectives were to improve healthcare accessibility, quality and equity in our community during the novel coronavirus pandemic. While initially focused on SARS-CoV-2 testing, our program quickly evolved to include preventive health services. The MHU program began as a location-based SARS-CoV-2 testing strategy coordinated with local and state public health agencies. Community needs motivated further program expansion to include additional preventive healthcare and social services. MHU deployment was targeted to disease “hotspots” based on publicly available SARS-CoV-2 testing data and community-level information about social vulnerability. This formative evaluation explores whether our MHU deployment strategy enabled us to reach patients from communities with heightened social vulnerability as intended. From 3/20/20-3/24/21, the Detroit MHU program reached a total of 32,523 people. The proportion of patients who resided in communities with top quartile Centers for Disease Control and Prevention Social Vulnerability Index rankings increased from 25% during location-based “drive-through” SARS-CoV-2 testing (3/20/20-4/13/20) to 27% after pivoting to a mobile platform (4/13/20-to-8/31/20; p = 0.01). The adoption of a data-driven deployment strategy resulted in further improvement; 41% of the patients who sought MHU services from 9/1/20-to-3/24/21 lived in vulnerable communities (Cochrane Armitage test for trend, p<0.001). Since 10/1/21, 1,837 people received social service referrals and, as of 3/15/21, 4,603 were administered at least one dose of COVID-19 vaccine. Our MHU program demonstrates the capacity to provide needed healthcare and social services to difficult-to-reach populations from areas with heightened social vulnerability. This model can be expanded to meet emerging pandemic needs, but it is also uniquely capable of improving health equity by addressing longstanding gaps in primary care and social services in vulnerable communities.

Research on the Deployment Strategy of Blockchain Nodes in the Agricultural Product Blockchain Traceability System

The agricultural product traceability system based on blockchain can monitor the entire growth cycle of agricultural products, trace it at any time, and cannot tamper with information without authorization. However, the energy consumption of the entire system is relatively high due to the introduction of blockchain technology. In order to alleviate the problem of high overall energy consumption, we are trying to reduce the communication power consumption between terminal sensors and the full nodes of the blockchain. We use the K-means algorithm, the DBSCAN algorithm and the improved DK fusion algorithm we proposed to deploy blockchain full nodes to the agricultural products sensors that have been determined to reduce the communication power consumption of the sensor terminals and improve the coverage of the full nodes to the terminal sensors.

A Multiphase Dynamic Deployment Mechanism of Virtualized Honeypots Based on Intelligent Attack Path Prediction

As an important deception defense method, a honeypot can be used to enhance the network’s active defense capability effectively. However, the existing rigid deployment method makes it difficult to deal with the uncertain strategic attack behaviors of the attackers. To solve such a problem, we propose a multiphase dynamic deployment mechanism of virtualized honeypots (MD2VH) based on the intelligent attack path prediction method. MD2VH depicts the attack and defense characteristics of both attackers and defenders through the Bayesian state attack graph, establishes a multiphase dynamic deployment optimization model of the virtualized honeypots based on the extended Markov’s decision-making process, and generates the deployment strategies dynamically by combining the online and offline reinforcement learning methods. Besides, we also implement a prototype system based on software-defined network and virtualization container, so as to evaluate the effectiveness of MD2VH. Experiments results show that the capture rate of MD2VH is maintained at about 90% in the case of both simple topology and complex topology. Compared with the simple intelligent deployment strategy, such a metric is increased by 20% to 60%, and the result is more stable under different types of the attacker’s strategy.