EMDL '17Pub Date : 2017-06-23DOI: 10.1145/3089801.3089806
Jia Guo, M. Potkonjak
{"title":"Pruning Filters and Classes: Towards On-Device Customization of Convolutional Neural Networks","authors":"Jia Guo, M. Potkonjak","doi":"10.1145/3089801.3089806","DOIUrl":"https://doi.org/10.1145/3089801.3089806","url":null,"abstract":"In recent years, we have witnessed more and more mobile applications based on deep learning. Widely used as they may be, those applications provide little flexibility to cater to the diversified needs of different groups of users. For users facing a classification problem, it is natural that some classes are more important to them, while the rest are not. We thus propose a lightweight method that allows users to prune the unneeded classes together with associated filters from convolutional neural networks (CNNs). Such customization can result in substantial reduction in computational costs at test time. Early results have shown that after pruning the Network-in-Network (NIN) model on CIFAR-10 datasetcite{lim2013network} down to a 5-class classifier, we can trade a 3% loss in accuracy for a 1.63$times$ gain in energy consumption and a 1.24$times$ improvement in latency when experimenting on an off-the-shelf smartphone, while the procedure incurs with very little overhead. After pruning, the custom-tailored model can still achieve a higher classification accuracy than the unmodified classifier because of a smaller problem space that more accurately reflects users' needs.","PeriodicalId":125567,"journal":{"name":"EMDL '17","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129098440","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
EMDL '17Pub Date : 2017-06-23DOI: 10.1145/3089801.3089803
Beomjun Kim, Yongsu Jeon, Heejin Park, Dongheon Han, Yunju Baek
{"title":"Design and Implementation of the Vehicular Camera System using Deep Neural Network Compression","authors":"Beomjun Kim, Yongsu Jeon, Heejin Park, Dongheon Han, Yunju Baek","doi":"10.1145/3089801.3089803","DOIUrl":"https://doi.org/10.1145/3089801.3089803","url":null,"abstract":"In recent years, there is a growing interest in advanced driver assistance systems, which can reduce the risk of accidents on various roads. Many vehicular technologies use camera information to collect roadside information. But research and development of image recognition in embedded environments is challenging. Therefore, there is a requirement for research to apply open source image recognition technology to embedded platform. Deep learning, a technology that has been on the spotlight recently, shows excellent performance in image recognition. However, deep learning has a problem that the network size and amount of computation are too large. In this paper, we design and implement a deep learning based object recognition system to recognize vehicles on the road. We recognize vehicles using Faster-RCNN, which has excellent object recognition capabilities. However, it is problematic to apply it to an embedded device due to the feature of deep learning. Therefore, we propose and evaluate the performance of object recognition with quantization and pruning. Through the evaluation, we will show that the proposed system reduces the network size to 16% and reduces the operating time to 64% without a significant decrease in recognition accuracy.","PeriodicalId":125567,"journal":{"name":"EMDL '17","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120978613","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
EMDL '17Pub Date : 2017-06-03DOI: 10.1145/3089801.3089804
Qingqing Cao, Niranjan Balasubramanian, A. Balasubramanian
{"title":"MobiRNN: Efficient Recurrent Neural Network Execution on Mobile GPU","authors":"Qingqing Cao, Niranjan Balasubramanian, A. Balasubramanian","doi":"10.1145/3089801.3089804","DOIUrl":"https://doi.org/10.1145/3089801.3089804","url":null,"abstract":"In this paper, we explore optimizations to run Recurrent Neural Network (RNN) models locally on mobile devices. RNN models are widely used for Natural Language Processing, Machine translation, and other tasks. However, existing mobile applications that use RNN models do so on the cloud. To address privacy and efficiency concerns, we show how RNN models can be run locally on mobile devices. Existing work on porting deep learning models to mobile devices focus on Convolution Neural Networks (CNNs) and cannot be applied directly to RNN models. In response, we present MobiRNN, a mobile-specific optimization framework that implements GPU offloading specifically for mobile GPUs. Evaluations using an RNN model for activity recognition shows that MobiRNN does significantly decrease the latency of running RNN models on phones.","PeriodicalId":125567,"journal":{"name":"EMDL '17","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128889941","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
EMDL '17Pub Date : 2017-05-17DOI: 10.1145/3089801.3089802
Kleomenis Katevas, Ilias Leontiadis, M. Pielot, J. Serrà
{"title":"Practical Processing of Mobile Sensor Data for Continual Deep Learning Predictions","authors":"Kleomenis Katevas, Ilias Leontiadis, M. Pielot, J. Serrà","doi":"10.1145/3089801.3089802","DOIUrl":"https://doi.org/10.1145/3089801.3089802","url":null,"abstract":"We present a practical approach for processing mobile sensor time series data for continual deep learning predictions. The approach comprises data cleaning, normalization, capping, time-based compression, and finally classification with a recurrent neural network. We demonstrate the effectiveness of the approach in a case study with 279 participants. On the basis of sparse sensor events, the network continually predicts whether the participants would attend to a notification within 10 minutes. Compared to a random baseline, the classifier achieves a 40% performance increase (AUC of 0.702) on a withheld test set. This approach allows to forgo resource-intensive, domain-specific, error-prone feature engineering, which may drastically increase the applicability of machine learning to mobile phone sensor data.","PeriodicalId":125567,"journal":{"name":"EMDL '17","volume":"79 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114299636","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}