xDNN: Inference for Deep Convolutional Neural Networks

We present xDNN, an end-to-end system for deep-learning inference based on a family of specialized hardware processors synthesized on Field-Programmable Gate Array (FPGAs) and Convolution Neural Networks (CNN). We present a design optimized for low latency, high throughput, and high compute efficiency with no batching. The design is scalable and a parametric function of the number of multiply-accumulate units, on-chip memory hierarchy, and numerical precision. The design can produce a scale-down processor for embedded devices, replicated to produce more cores for larger devices, or resized to optimize efficiency. On Xilinx Virtex Ultrascale+ VU13P FPGA, we achieve 800 MHz that is close to the Digital Signal Processing maximum frequency and above 80% efficiency of on-chip compute resources. On top of our processor family, we present a runtime system enabling the execution of different networks for different input sizes (i.e., from 224× 224 to 2048× 1024). We present a compiler that reads CNNs from native frameworks (i.e., MXNet, Caffe, Keras, and Tensorflow), optimizes them, generates codes, and provides performance estimates. The compiler combines quantization information from the native environment and optimizations to feed the runtime with code as efficient as any hardware expert could write. We present tools partitioning a CNN into subgraphs for the division of work to CPU cores and FPGAs. Notice that the software will not change when or if the FPGA design becomes an ASIC, making our work vertical and not just a proof-of-concept FPGA project. We show experimental results for accuracy, latency, and power for several networks: In summary, we can achieve up to 4 times higher throughput, 3 times better power efficiency than the GPUs, and up to 20 times higher throughput than the latest CPUs. To our knowledge, we provide solutions faster than any previous FPGA-based solutions and comparable to any other top-of-the-shelves solutions.

THE ANALYSIS OF COMPLIMENT STRATEGIES IN A MOVIE ENTITLED THE SPECTACULAR NOW

<p>This research focuses on what kind of compliment and compliment responses in a movie entitled <em>The Spectacular Now</em>. This paper tries to find out the types of compliment’s topics, to reveal the functions of each type of compliment’s topics and to figure out the types of compliment responses employed by the characters in the movie entitled <em>The Spectacular Now</em>. This research is descriptive qualitative research. This research uses a pragmatic approach. The technique used is purposive sampling, meaning that the researcher took all of dialogues containing giving and responding compliments as data. The results show that first the most common topic of compliment used by the characters is the ability topic. Following the Wolfson’s theory, there are five functions found in the movie. They are to express admiration or approval, to maintain rapport, to modify sarcasm, to encourage, and to give positive evaluation for someone. Furthermore, there are eight compliment responses found in the movie based on Herbert’s classification. They are appreciation token, no-acknowledgement, comment acceptance, question, scale down, disagreement, request, and return. Based on the analysis, the most widely used is appreciation token with verbal acceptance.</p>

Microfluidic single-cell scale-down bioreactors: A proof-of-concept for the growth of Corynebacterium glutamicum at oscillating pH values

In large-scale bioreactors, gradients in cultivation parameters such as oxygen, substrate and pH result in fluctuating environments. pH fluctuations were identified as a critical parameter for bioprocess performance. Traditionally, scale-down systems at the laboratory scale are used to analyze the effects of fluctuating pH values on strain and thus process performance. Here, we demonstrate the application of dynamic microfluidic single-cell cultivation (dMSCC) as a novel scale-down system for the characterization of Corynebacterium glutamicum growth using oscillating pH conditions as a model parameter. A detailed comparison between two-compartment reactor (two-CR) scale-down experiments and dMSCC was performed for one specific pH oscillation between reference pH 7 (~8 min) and disturbed pH 6 (~2 min). Similar reductions in growth rates were observed in both systems (dMSCC 21% and two-CR 27%). Afterward, systematic experiments at different symmetric and asymmetric pH oscillations between pH ranges of 4 −6 and 8 −11 and different intervals from 1 minute to 20 minutes, were performed to demonstrate the unique application range and throughput of the dMSCC system. Finally, the strength of the dMSCC application was demonstrated by mimicking fluctuating environmental conditions within large-scale bioprocesses, which is difficult to conduct using two-CRs.

“THANK YOU, IT REALLY MAKES MY DAY”: COMPLIMENT RESPONSES REVISITED

Complimenting is a typical speaking act and the method in which it is responded to can vary based on the culture of the speakers as well as the influence of other circumstances. The purpose of this study was to provide a more in-depth knowledge of compliment response research based on how it is used by university students learning EFL in everyday life. Furthermore, this study also aims at finding out if exposure to another culture affects university students learning English as a second language while responding to compliments. The data were garnered using data elicitation method by complimenting the participants’ look, possession, character, or aptitude. The finding found that the participants used ten types of responses; listed from the most frequently used type of compliment responses to the least used type of compliment responses: Comment Acceptance (8), Appreciation Token (2), Comment History (2), Question (2), Praise Upgrade (1), Reassignment (1), Return (1), Scale Down (1), Disagreement (1), and Qualification (1). According to the data, the majority of students in an international English literature class at Brawijaya University are likely to accept the compliments. Students tend to take compliments by thanking them and then making related comments. English-speaking countries consider a simple "thank you" to be an adequate response to a praise. This present study also confirms that short term encounterance with foreign culture exposure slightly affect EFL learners’ way of responding to compliments. This shows that the students learn the English language culture in terms of compliment.

Heavy metals in contaminated soil: source, accumulation, health risk and remediation process

Heavy metals contamination of soil is an issue of global concern that ultimately results in toxicity and diseases in humans and animals through consumption of food crops from contaminated soil. The toxic effects of these metals, even though they do not have any biological role, remain present in some or the other form harmful for the human body and its proper functioning. Heavy metals in the soil refers to some significant heavy metals of biological toxicity, including Cobalt (Co), Cadmium (Cd), Lead (Pb), Chromium (Cr), and Arsenic (As) etc. With the development of the global economy, both type and content of heavy metals in the soil caused by indiscriminate use for human purposes has altered their geochemical cycles and biochemical balance. There are many known sources of harmful metals, including the earth, which releases them into food, air, and water, and anthropogenic activities, such as the application of fertilizer in agriculture, the use of pesticides and herbicides, and irrigation. Other sources are automobile emissions, paints, cigarette smoking, industries, and sewage and waste disposal. This review gives details about some heavy metals their toxicity, bioaccumulation, biodegradation, mobility and solubility in contaminated soil along with their health effects. Remediation of heavy metal in contaminated soils is necessary to reduce the associated risks, make the available soil safe for agricultural production, enhance food security and scale down land tenure problems arising from changes in the land use pattern the review recommends regular monitoring of heavy metals in soil, vegetables and foodstuffs to prevent excessive accrual in food chain.

Binding Reactions at Finite Systems

A perpetual yearn exists among computational scientists to scale-down the size of physical systems, a desire shared as well with experimentalists able to track single molecules. A question then arises whether averages observed at small systems are the same as those observed at large, or macroscopic, systems. Utilizing statistical-mechanics formulations in ensembles in which the total numbers of particles are fixed, we demonstrate that system's properties of binding reactions are not homogeneous functions. That means averages of intensive properties, such as the concentration of the bound-state, at finite-systems are different than those at large-systems. The discrepancy increases with decreasing numbers of particles, temperature, and volume. As perplexing as it may sound, despite these variations in average quantities, extracting the equilibrium constant from systems of different sizes does yield the same value. The reason is that correlations in reactants' concentrations are ought be accounted for in the expression of the equilibrium constant, being negligible at large-scale but significant at small-scale. Similar arguments pertain to the calculations of the reaction rate-constants, more specifically, the bimolecular rate of the forward reaction is related to the average of the product (and not to the product of the averages) of the reactants' concentrations. Furthermore, we derive relations aiming to predict the composition of the system only from the value of the equilibrium constant. All predictions are validated by Monte-Carlo and molecular dynamics simulations. An important significance of these findings is that the expression of the equilibrium constant at finite systems is not dictated solely by the chemical equation but requires knowledge of the elementary processes involved.

Evaluating the resilience of a full-scale down-flow hanging sponge reactor to shock-loadings

The effect of shock-loading on the performance of a full-scale down-flow hanging sponge (DHS), used to treat effluent from an up-flow anaerobic sludge blanket treating municipal sewage, was evaluated. This full-scale DHS reactor comprises a polyurethane sponge that retains the sludge. It has a capacity of 500 m3/day and, at the time of this study, had been operating at a sewage treatment plant in India for more than 1,300 days. The DHS reactor was exposed to shock-loadings of organics at double the normal rate for 400 min under summer and winter conditions. The results showed that the DHS reactor maintained stable operation under the organic shock-loading and that it returned to a steady state soon after restart, confirming that the reactor was resilient to organic shock-loadings.

A Quasi-Z-Source-Based Inductive Power Transfer System for Constant Current/Constant Voltage Charging Applications

This article proposes a quasi-Z-source (qZS)-based Inductive Power Transfer (IPT) system for Electric Vehicles (EVs) charging applications. The IPT systems use the magnetic field to transfer power between two coils wirelessly, achieving improved reliability, safety and less environmental impact. Compared to the conventional IPT system, the proposed qZS-IPT system simultaneously achieves DC/DC regulation and DC/AC conversion through a single-stage conversion, thus lowering the cost and complexity of the system. Moreover, the reliability of the system is improved thanks to the qZS network shoot-though immunity and the reduced number of switches. To ensure the battery efficient charging and long service life, the constant current/constant voltage (CC/CV) method is considered. With the proposed innovative modulation scheme, the qZS can easily change between buck and boost modes, respectively, lowering or increasing the secondary side current. A theoretical analysis is presented for system design. Simulation results based on a 25 kW (200 V/135 A) low duty EV charger are presented to verify the effectiveness of the proposed scheme. Experimental tests are performed on a 150 W scale-down prototype to validate the analysis and demonstrate the effectiveness of the proposed qZS-IPT system for CC/CV chargers.