<p>In recent years, deep learning models have become popular in the real-time embedded application, but there are many complexities for hardware deployment because of limited resources such as memory, computational power, and energy. Recent research in the field of deep learning focuses on reducing the model size of the Convolution Neural Network (CNN) by various compression techniques like Architectural compression, Pruning, Quantization, and Encoding (e.g., Huffman encoding). Network pruning is one of the promising technique to solve these problems.</p> <p>This thesis proposes methods to prune the convolution neural network (SqueezeNet) without introducing network sparsity in the pruned model. </p> <p>This thesis proposes three methods to prune the CNN to decrease the model size of CNN without a significant drop in the accuracy of the model.</p> <p>1: Pruning based on Taylor expansion of change in cost function Delta C.</p> <p>2: Pruning based on L₂ normalization of activation maps.</p> <p>3: Pruning based on a combination of method 1 and method 2.</p><p>The proposed methods use various ranking methods to rank the convolution kernels and prune the lower ranked filters afterwards SqueezeNet model is fine-tuned by backpropagation. Transfer learning technique is used to train the SqueezeNet on the CIFAR-10 dataset. Results show that the proposed approach reduces the SqueezeNet model by 72% without a significant drop in the accuracy of the model (optimal pruning efficiency result). Results also show that Pruning based on a combination of Taylor expansion of the cost function and L₂ normalization of activation maps achieves better pruning efficiency compared to other individual pruning criteria and most of the pruned kernels are from mid and high-level layers. The Pruned model is deployed on BlueBox 2.0 using RTMaps software and model performance was evaluated.</p><p></p>

Indiana University-Purdue University Indianapolis (IUPUI) / In recent years, deep learning models have become popular in the real-time embedded application, but there are many complexities for hardware deployment because of limited resources such as memory, computational power, and energy. Recent research in the field of deep learning focuses on reducing the model size of the Convolution Neural Network (CNN) by various compression techniques like Architectural compression, Pruning, Quantization, and Encoding (e.g., Huffman encoding). Network pruning is one of the promising technique to solve these problems. This thesis proposes methods to prune the convolution neural network (SqueezeNet) without introducing network sparsity in the pruned model. This thesis proposes three methods to prune the CNN to decrease the model size of CNN without a significant drop in the accuracy of the model. 1: Pruning based on Taylor expansion of change in cost function Delta C. 2: Pruning based on L2 normalization of activation maps. 3: Pruning based on a combination of method 1 and method 2. The proposed methods use various ranking methods to rank the convolution kernels and prune the lower ranked filters afterwards SqueezeNet model is fine-tuned by backpropagation. Transfer learning technique is used to train the SqueezeNet on the CIFAR-10 dataset. Results show that the proposed approach reduces the SqueezeNet model by 72% without a significant drop in the accuracy of the model (optimal pruning efficiency result). Results also show that Pruning based on a combination of Taylor expansion of the cost function and L2 normalization of activation maps achieves better pruning efficiency compared to other individual pruning criteria and most of the pruned kernels are from mid and high-level layers. The Pruned model is deployed on BlueBox 2.0 using RTMaps software and model performance was evaluated.

Many variations of algorithms for finding the shortest path in a large graph have been introduced recently due to the needs of applications like the Geographic Information System (GIS) or Intelligent Transportation System (ITS). The primary subjects of those algorithms are materialization and hierarchical path views. Some studies focus on the materialization and sacrifice the pre-computational costs and storage costs for faster computation of a query. Other studies focus on the shortest-path algorithm, which has less pre-computation and storage but takes more time to compute the shortest path. The main objective of this thesis is to accelerate the computation time for the shortest-path queries while keeping the degree of materialization as low as possible. This thesis explores two different categories: 1) the reduction of the I/O-costs for multiple queries, and 2) the reduction of search spaces in a graph. The thesis proposes two simple algorithms to reduce the I/O-costs, especially for multiple queries. To tackle the problem of reducing search spaces, we give two different levels of materializations, namely, the <i>boundary set distance matrix</i> and <i>x-Hop sketch graph</i>, both of which materialize the shortest-path view of the boundary nodes in a partitioned graph. Our experiments show that a combination of the suggested solutions for 1) and 2) performs better than the original Disk-based SP algorithm [7], on which our work is based, and requires much less storage than <i>HEPV</i> [3].

A pluridade de características de espécies nativas, aliada a leis que limitam sua exploração, acarreta na necessidade de estabelecimento de novas espécies capazes de suprir a demanda madeireira brasileira. Os Eucalyptus spp. são altamente produtivos sob clima tropical, sendo extensamente plantados no país para diversas finalidades, em especial papel e celulose, energia, painéis e serraria, entretanto, pouca ênfase tem sido dada a esta última, devido ao horizonte longo de planejamento, dificuldades no processamento primário e defeitos na madeira apresentados pelas espécies mais produtivas do gênero. Visando traçar estratégias que minimizem essas dificuldades, analisamos o efeito de 2 formas de manipulação de copa - destopo e desrama -, sob diferentes regimes de luminosidade, em parâmetros da copa e do tronco em híbridos de Eucalyptus urophylla dos 11 aos 23 meses de idade. Tanto o destopo quanto a desrama afetaram negativamente o incremento diamétrico e volumétrico do híbrido, entretanto, árvores destopadas sob níveis maiores de luminosidade exibiram fatores de afilamento menores, quando comparadas ao tratamento Controle. Dentro do mesmo regime de luminosidade, árvores desramadas também apresentaram fatores de afilamento inferiores. Apesar de sua alta demanda por luz, o híbrido foi capaz de alterar sua estrutura foliar de acordo com o nível de luminosidade: árvores sob regimes de luminosidade reduzida adaptaram suas folhas à sombra (folhas de sombra), enquanto que àquelas sem limitação de luminosidade folhas adaptadas ao sol (folhas de sol). Brotações em resposta ao destopo se resumiram à brotações epicórmicas próximo à seção destopada. / The diversity of characteristics of native species, allied to laws that limit their exploration, reinforces the need of establishing news species to supply the national wood demand. The Eucalyptus spp. are highly productive under tropical whether, being extensively planted in the country to many finalities, especially paper, pulp, energy, panels and sawn wood, however, low attention is being given to the last due to the long term planning, difficulties in the primary processing and wood blemishes present in the most productive species of the gender. Aiming to trace strategies that minimize these hardships, we analyzed the effects of two forms of crown manipulation - topping and pruning -, under different light patterns, in parameters of crown and bole of Eucalytpus urophylla hybrids from 11 to 23 months of age. Both topping and pruning impacted negatively the diametric and volumetric increment, although topped trees showed smaller taper when compared to the Control treatment. Within the same light pattern pruned trees also presented smaller taper. Although the hybrid demands high amounts of light, it was capable of altering its foliage structure according to the light level: trees under reduced light adapted its leaves to shadow (\"shadow leaves\"), while those without light limitation leaves adapted to sun (\"sunleaves\"). Shoots in response to topping were resumed to epicormic shoots close to the topped section.

A pluridade de características de espécies nativas, aliada a leis que limitam sua exploração, acarreta na necessidade de estabelecimento de novas espécies capazes de suprir a demanda madeireira brasileira. Os Eucalyptus spp. são altamente produtivos sob clima tropical, sendo extensamente plantados no país para diversas finalidades, em especial papel e celulose, energia, painéis e serraria, entretanto, pouca ênfase tem sido dada a esta última, devido ao horizonte longo de planejamento, dificuldades no processamento primário e defeitos na madeira apresentados pelas espécies mais produtivas do gênero. Visando traçar estratégias que minimizem essas dificuldades, analisamos o efeito de 2 formas de manipulação de copa - destopo e desrama -, sob diferentes regimes de luminosidade, em parâmetros da copa e do tronco em híbridos de Eucalyptus urophylla dos 11 aos 23 meses de idade. Tanto o destopo quanto a desrama afetaram negativamente o incremento diamétrico e volumétrico do híbrido, entretanto, árvores destopadas sob níveis maiores de luminosidade exibiram fatores de afilamento menores, quando comparadas ao tratamento Controle. Dentro do mesmo regime de luminosidade, árvores desramadas também apresentaram fatores de afilamento inferiores. Apesar de sua alta demanda por luz, o híbrido foi capaz de alterar sua estrutura foliar de acordo com o nível de luminosidade: árvores sob regimes de luminosidade reduzida adaptaram suas folhas à sombra (folhas de sombra), enquanto que àquelas sem limitação de luminosidade folhas adaptadas ao sol (folhas de sol). Brotações em resposta ao destopo se resumiram à brotações epicórmicas próximo à seção destopada. / The diversity of characteristics of native species, allied to laws that limit their exploration, reinforces the need of establishing news species to supply the national wood demand. The Eucalyptus spp. are highly productive under tropical whether, being extensively planted in the country to many finalities, especially paper, pulp, energy, panels and sawn wood, however, low attention is being given to the last due to the long term planning, difficulties in the primary processing and wood blemishes present in the most productive species of the gender. Aiming to trace strategies that minimize these hardships, we analyzed the effects of two forms of crown manipulation - topping and pruning -, under different light patterns, in parameters of crown and bole of Eucalytpus urophylla hybrids from 11 to 23 months of age. Both topping and pruning impacted negatively the diametric and volumetric increment, although topped trees showed smaller taper when compared to the Control treatment. Within the same light pattern pruned trees also presented smaller taper. Although the hybrid demands high amounts of light, it was capable of altering its foliage structure according to the light level: trees under reduced light adapted its leaves to shadow (\"shadow leaves\"), while those without light limitation leaves adapted to sun (\"sunleaves\"). Shoots in response to topping were resumed to epicormic shoots close to the topped section.

Atualmente a tomaticultura brasileira ostenta o título de cultivo com o maior custo de produção médio no país. Dentre algumas alternativas a fim de minimizar os elevados gastos está o uso de híbridos de crescimento semideterminado em sistema de meia-estaca, devido a menor necessidade de podas, insumos e mão de obra. Realizaram-se dois experimentos, em condições de campo aberto, com os objetivos de (a) determinar o efeito de diferentes sistemas de poda (P1, P2 e P3) e distintos espaçamentos entre plantas (E1 e E2) sobre a produção e seus componentes para o híbrido Silvety, de hábito de crescimento semideterminado do segmento Salada Longa Vida (SLV) (Experimento 1); e (b) avaliar o efeito de distintos sistemas de poda (P1 e P2) sobre a produção e seus componentes para os híbridos Silvety, Sagatan e H003, de hábito de crescimento semideterminado do segmento SLV (Experimento 2). Em ambos os experimentos adotou-se o delineamento em blocos ao acaso com os tratamentos em esquema fatorial 3 x 2 e quatro repetições. O experimento 1 foi conduzido em Jacuí, MG, de maio de 2015 a novembro de 2015, já o experimento 2 foi conduzido em Holambra, SP, de fevereiro de 2016 a agosto de 2016. Avaliaram-se o número total de frutos (NTF), a produção total por parcela (PTP), o número de frutos comercializáveis (NFC), a produção comercial (PCO), a produção por planta (PPP), a massa média dos frutos (MMF), o número de frutos por classe AAA, AA e A (NAAA, NAA e NA), a massa de frutos por classe AAA, AA e A (MAAA, MAA e MA), o número de hastes (NH) o número de inflorescências (NI), o número de frutos por inflorescência (NFI), o número de frutos com rachaduras cuticulares (CUT), o número de frutos ocos (OCO) e o número de frutos manchados (MAN). Nas condições em que os experimentos foram conduzidos e de acordo com os resultados obtidos conclui-se que no experimento 1, a poda P2e o espaçamento E1 mostraram-se os mais indicados para o híbrido Silvety. No experimento 2, novamente, a poda P2 mostrou ser a mais recomendada para os híbridos Silvety, Sagatan e H003. / Currently, the Brazilian tomato crop holds the title with the highest average cost of production in the country. Among the alternatives to minimize the high expenses, is the use of hybrids with semi-determinate growth habit in a training system of \"half-stakes\", due to the less need of pruning, inputs and labor. Two experiments were carried out under open field conditions, aiming to (a) to determine the effect of different pruning system (P1, P2 and P3) and spacing between plants (E1 and E2) on the production and its components for the Silvety hybrid, of semi-determinate growth habit of varietal segment Salad Extended Shelf Life (SLV) (Experiment 1), and (b) to evaluate the effect of pruning system (P1 and P2) on the production and its components for the Silvety, Sagatan and H003 hybrids, of semi- determinate growth habit of segment SLV (Experiment 2). In both experiments it was adopted the randomized block design with the treatments in a 3 x 2 factorial scheme, and four replications. Experiment 1 was conducted in Jacuí, MG, from May 2015 to November 2015, and Experiment 2 was conducted in Holambra, SP, from February 2016 to August 2016. There were assessed the total number of fruits (NTF), the total production per plot (PTP), the number of commercial fruits (NFC), the commercial production (PCO), the production per plant (PPP), the average mass of fruits (MMF), the number of fruits per classes AAA,AA and A (NAAA, NAA and NA), the mass of fruits per classes AAA, AA and A (MAAA, MAA and MA), the number of stems (NH), the number of clusters (NI), number of fruits per cluster (NFI), number of fruits with micro-cracking (CUT), number of hollow fruits (OCO), number of fruits with blotchy (MAN). Under the conditions that the trials were conducted and according to the results obtained, it was concluded that in experiment 1, the pruning system P2 and E1 spacing were the most suitable for Silvety hybrid. In experiment 2, again, the pruning system P2 showed to be the most recommended for Silvety, Sagatan and H003 hybrids.

No description available.

Thesis(M.Sc)--University of Limpopo, 2010. / Bush tea (Athrixia phylicoides DC.) has received interest as another indigenous South African herbal tea with potential for commercialization. The success of commercialization of bush tea hinges on maintenance or enhancement of quality of bush tea as a herbal beverage. Herbal tea quality is one of the critical factors in commercialization that would determine the price of tea for local sale and export. A research was conducted to determine chemical profiles of bush tea at different phenological stages and as influenced by pruning and application of growth regulators. A trial to determine the quality of tea harvested at various phenological stages (namely new growth, older growth and whole plants) from wild and cultivated bush tea was conducted. This was to help determine the best phenological stage to harvest bush tea of best quality. In cultivated bush tea, harvested new growth or as whole plants proved to be of higher quality owing to their higher polyphenol and tannin attributes respectively. In wild bush tea, both new and older growth proved to be of good quality owing to the higher total polyphenol content and higher total antioxidants, respectively. A trial to determine the effect of pruning at different heights on growth and quality of bush tea was also conducted. Pruning of bush tea largely led to yield reduction. Unpruned tea plants remained the tallest plants, with higher number of branches, bigger leaf area and a larger biomass than apically, middle and base pruned bush tea plants. Pruning at different heights also proved to have little or no effect on quality of bush tea. While only total polyphenols remained higher in unpruned tea plants, no significant differences were observed in tannin and total antioxidant content in unpruned, apically, middle and base pruned tea plants. A third trial was conducted to investigate the effects of gibberellins on sprouting and quality of bush tea. The trial results showed that gibberellin application had a iv favourable effect on growth of bush tea, with application of 3% and 4% yielding the highest growth. The results also indicated a declining total polyphenol and antioxidant content with increasing gibberellin application rate, while tannins peaked at 2% application rate.

The tart cherry (Prunus cerasus) is the most significant fruit tree crop in Utah, accounting for roughly 51% of the total statewide commercial fruit acreage. In order to accommodate harvesting equipment, tart cherries are grown in conventional orchards with large trees spaced up to 5.5 meters apart. New methods of harvest are adapted to smaller trees in tighter spaced high density (HD) orchards. HD orchards bear fruit earlier in the orchard life than conventional orchards, but likely require different pruning and management strategies, which have not yet been determined for tart cherry. Experimental HD orchards were used to determine the type of renewal pruning cuts to maintain orchard productivity, and to determine whether mechanical pruning (hedging) could be used to maintain tree size. It was found that when removing branches during pruning, leaving the branch stub at least 10 cm long greatly increases the likelihood of getting adequate renewal growth. Mechanical hedging at bloom or 45 days after bloom did not change yields or fruit quality when applied to a well-pruned and maintained orchard. Spray pattern analysis in these canopies suggested that hedging creates a more consistent canopy density than unhedged HD canopies. This research provides distinct guidelines on renewal pruning of tart cherry to maintain productivity, and shows that mechanical hedging strategies can be a viable option for maintaining tree size in HD plantings without increasing pruning costs.

All practical text search systems use inverted indexes to quickly resolve user queries. Offline index construction algorithms, where queries are not accepted during construction, have been the subject of much prior research. As a result, current techniques can invert virtually unlimited amounts of text in limited main memory, making efficient use of both time and disk space. However, these algorithms assume that the collection does not change during the use of the index. This thesis examines the task of index maintenance, the problem of adapting an inverted index to reflect changes in the collection it describes. Existing approaches to index maintenance are discussed, including proposed optimisations. We present analysis and empirical evidence suggesting that existing maintenance algorithms either scale poorly to large collections, or significantly degrade query resolution speed. In addition, we propose a new strategy for index maintenance that trades a strictly controlled amount of querying efficiency for greatly increased maintenance speed and scalability. Analysis and empirical results are presented that show that this new algorithm is a useful trade-off between indexing and querying efficiency. In scenarios described in Chapter 7, the use of the new maintenance algorithm reduces the time required to construct an index to under one sixth of the time taken by algorithms that maintain contiguous inverted lists. In addition to work on index maintenance, we present a new technique for accumulator pruning during ranked query evaluation, as well as providing evidence that existing approaches are unsatisfactory for collections of large size. Accumulator pruning is a key problem in both querying efficiency and overall text search system efficiency. Existing approaches either fail to bound the memory footprint required for query evaluation, or suffer loss of retrieval accuracy. In contrast, the new pruning algorithm can be used to limit the memory footprint of ranked query evaluation, and in our experiments gives retrieval accuracy not worse than previous alternatives. The results presented in this thesis are validated with robust experiments, which utilise collections of significant size, containing real data, and tested using appropriate numbers of real queries. The techniques presented in this thesis allow information retrieval applications to efficiently index and search changing collections, a task that has been historically problematic.