To reduce memory footprint and run-time latency, techniques such as neural network pruning and binarization have been explored separately. However, it is unclear how to combine the best of the two worlds to get extremely small and efficient models. In this paper, we, for the first time, define the filter-level pruning problem for binary neural networks, which cannot be solved by simply migrating existing structural pruning methods for full-precision models. A novel learning-based approach is proposed to prune filters in our main/subsidiary network framework, where the main network is responsible for learning representative features to optimize the prediction performance, and the subsidiary component works as a filter selector on the main network. To avoid gradient mismatch when training the subsidiary component, we propose a layer-wise and bottom-up scheme.We also provide the theoretical and experimental comparison between our learning-based and greedy rule-based methods.Finally, we empirically demonstrate the effectiveness of our approach applied on several binary models, including binarized NIN, VGG-11, and ResNet-18, on various image classification datasets. For binary ResNet-18 on ImageNet, we use 78.6% filters but can achieve slightly better test error 49.87% (50.02%-0.15%) than the original model.1 Fully connected layers can be implemented as convolution. Therefore, in the rest of the paper, we mainly focus on convolutional layers.
With the rapid development of urban modernization, the pile anchor support system has been widely used in the deep foundation pit project. Because the deformation of the pile anchor support system directly affects the overall stability of the foundation pit and brings serious potential safety accidents to the construction of the deep foundation pit, avoiding the deformation of the pile anchor support system can effectively prevent the occurrence of the construction safety accidents of the deep foundation pit. Based on many years of construction experience, the deformation rule of pile anchor support system of deep foundation pit is taken as the research object. This paper starts from the scheme design of pile anchor support system and the monitoring data of the foundation pit, and combines with the actual deformation of the pile anchor support system in the construction site, summarizes the causes and prevention measures of the deformation of the pile anchor support system in the construction stage. The research results show that in the initial stage of construction, the top of the pile anchor support system inclines forward to the pit and gradually changes into a bow shape. This kind of deformation is related to the increase of excavation depth and the erosion of active soil pressure area by surface water. In the later stage of construction, the deformation of the top of the foundation pit converges and the bottom of the pile anchor support system protrudes into the pit, which is related to the erosion of the passive soil pressure area by rainwater and the embedded depth of the pile foundation.
With the rapid development of urbanization in China, people’s requirements for the quality of life and environment are constantly improving. However, the shortage of resources and the deterioration of environment are becoming more and more serious. Environmental protection has become the only way for the sustainable development of society. Therefore, the application of more and more new green building materials has become the key to the sustainable development of the construction industry. In the face of the rapid development of modern construction industry, engineering quality and functional requirements have become the basic requirements, the green environmental protection has become a higher requirement. The application of green environmental protection building materials in engineering construction is the main development direction of the transformation of modern construction industry. At the same time, it conforms to the policy of environmentally sustainable development of all sectors of society, and promotes the construction industry to meet the higher requirements of green environmental protection living environment. This paper studied the performance and waterproof mechanism of a kind of high polymer (HDPE) self-adhered waterproofing membrane and its application in construction. Compared with the traditional waterproofing membrane, the high polymer self-adhered waterproofing membrane can give full play to the advantages of environmental protection on the premise of meeting the building quality.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.