Search

Banana stem is a known alternative to wood for the production of pulp and paper. During the production of paper, it is extremely important to save costs and time while at the same time enhancing quality of produced paper. In the present study, paper was developed from banana stem fiber by optimizing its density. Response surface methodology (RSM) using Box-Behnken design (BBD) was used to determine optimum blending parameters of the developed papers. The influence of banana stem fiber amount (250-1000 g), water amount (1-2liters) and blending time (5-10 min) on paper density was ana)lyzed using software Stat-Ease 360, Version 13. The optimum conditions of banana stem fiber amount, water amount and blending time to achieve a density of 675.75 g/m3 were determined as 461.83 g, 1.31liters, and 8.53 min, respectively. ANOVA results revealed that the quadratic model best fits the density response, with blending time and banana stem fiber amount as the most significant model terms. Linear effects were more predominant than quadratic and interaction effects. The developed density model was in close agreement with experimental findings with an R² value of 0.9748. Developed paper had cellulosic contents, with volatiles as high as 82.11%wb and 2.11% wb ash composition. Future works on thermal properties of paper are encouraged. [ABSTRACT FROM AUTHOR]

With increasingly complex neural network architectures and heterogeneous device characteristics, finding a reasonable graph partitioning and device placement strategy is challenging. There have been prior attempts at learned approaches for solving device placement, these approaches are computationally expensive, unable to handle large graphs consisting over 50000 nodes, and do not generalize well to unseen graphs. To address all these limitations, we propose an efficient single-shot, generalized deep RL method (SGDP) based on a scalable sequential attention mechanism over a graph neural network that is transferable to new graphs. On a diverse set of representative deep learning models, our method on average achieves 20% improvement over human placement and 18% improvement over the prior art with 15× faster convergence. We are the first to demonstrate super human performance on 8-layer recurrent neural network language model and 8-layer GNMT consisting of over 50000 nodes, on 8-GPUs. We provide rationales and sensitivity study on model architecture selections. [ABSTRACT FROM AUTHOR]

There has been a growing interest in laser cleaning applications for a variety of organic materials such as paper, parchment, textiles, and leather during the last decade. However, archaeological organic materials, notably papyrus, have rarely been investigated. This contribution examines whether removal of burial encrustation can be justified in view of its short-term and long-term effects on the substrate. To examine this, tests using mock objects have been performed. Using artificially soiled and archaeological papyrus samples, optimization of laser cleaning parameters using a picosecond laser (1064 nm, various operating conditions) was attempted. Optimization was based on colorimetry, optical microscopy, scanning electron microscopy, Fourier transform infrared spectroscopy, and cellulose degree of polymerization data, both before and after accelerated degradation. In papyrus, there is no clear damage threshold, and substrate degradation can always be observed and is comparable in treated (cleaned) and untreated (soiled) objects. Therefore, the decision on whether to clean papyrus using lasers is predominantly based on aesthetic and treatability (e.g. need for consolidation) criteria. [ABSTRACT FROM AUTHOR]

Abstract: The overproduce-and-choose strategy, which is divided into the overproduction and selection phases, has traditionally focused on finding the most accurate subset of classifiers at the selection phase, and using it to predict the class of all the samples in the test data set. It is therefore, a static classifier ensemble selection strategy. In this paper, we propose a dynamic overproduce-and-choose strategy which combines optimization and dynamic selection in a two-level selection phase to allow the selection of the most confident subset of classifiers to label each test sample individually. The optimization level is intended to generate a population of highly accurate candidate classifier ensembles, while the dynamic selection level applies measures of confidence to reveal the candidate ensemble with the highest degree of confidence in the current decision. Experimental results conducted to compare the proposed method to a static overproduce-and-choose strategy and a classical dynamic classifier selection approach demonstrate that our method outperforms both these selection-based methods, and is also more efficient in terms of performance than combining the decisions of all classifiers in the initial pool. [Copyright &y& Elsevier]