Filters








418,985 Hits in 7.4 sec

On the efficient representation and execution of deep acoustic models [article]

Raziel Alvarez, Rohit Prabhavalkar, Anton Bakhtin
2016 arXiv   pre-print
We validate the proposed techniques by applying them to a long short-term memory-based acoustic model on an open-ended large vocabulary speech recognition task.  ...  The proposed quantization scheme leads to significant memory savings and enables the use of optimized hardware instructions for integer arithmetic, thus significantly reducing the cost of inference.  ...  Conclusions We propose a simple and computationally efficient quantization scheme for training and execution of deep learning models.  ... 
arXiv:1607.04683v2 fatcat:uquxbffaajawxosb6a6sn7ylye

Modeling and Simulation of the Deep-sea Mining Vehicle's Hydraulic Execution System

Yun LIU, Yong JIANG, Yang LI, Yong-Jie ZHANG
2017 DEStech Transactions on Engineering and Technology Research  
Take the deep-sea mining vehicle's hydraulic execution system for research object.  ...  Eatablish its simulation model and simulate various extreme obstacle conditions in deep sea environment using software AMESim.  ...  Acknowledgement This work was financially supported by Scientific research projects of Weifang University (2012Z08).  ... 
doi:10.12783/dtetr/mdm2016/4991 fatcat:5odnvodmrbhqrkrjpo3hsk3e2q

Modeling and Simulation of the Deep-sea Mining Vehicle's Hydraulic Execution System

Yun LIU, Yong JIANG, Yang LI, Yong-Jie ZHANG
2017 DEStech Transactions on Engineering and Technology Research  
Take the deep-sea mining vehicle's hydraulic execution system for research object.  ...  Eatablish its simulation model and simulate various extreme obstacle conditions in deep sea environment using software AMESim.  ...  Acknowledgement This work was financially supported by Scientific research projects of Weifang University (2012Z08).  ... 
doi:10.12783/dtetr/mdm2016/5006 fatcat:4v46mmt2cjhazcdcci5vichpnu

Predicting the Computational Cost of Deep Learning Models

Daniel Justus, John Brennan, Stephen Bonner, Andrew Stephen McGough
2018 2018 IEEE International Conference on Big Data (Big Data)  
In this work we propose an alternative approach in which we train a deep learning network to predict the execution time for parts of a deep learning network.  ...  Such as the time to load data from memory or loss of performance due to non-optimal parallel execution.  ...  ACKNOWLEDGMENT We would like to thank Innovate UK and the European Regional Development Fund for their contribution towards making this work possible.  ... 
doi:10.1109/bigdata.2018.8622396 dblp:conf/bigdataconf/JustusBBM18 fatcat:owsmwkuvgrhztehip52mc66an4

Predicting the Computational Cost of Deep Learning Models [article]

Daniel Justus, John Brennan, Stephen Bonner, Andrew Stephen McGough
2018 arXiv   pre-print
In this work we propose an alternative approach in which we train a deep learning network to predict the execution time for parts of a deep learning network.  ...  Such as the time to load data from memory or loss of performance due to non-optimal parallel execution.  ...  ACKNOWLEDGMENT We would like to thank Innovate UK and the European Regional Development Fund for their contribution towards making this work possible.  ... 
arXiv:1811.11880v1 fatcat:ou4zbcf36naerjhzt3kxvd2ck4

An Early Resource Characterization of Deep Learning on Wearables, Smartphones and Internet-of-Things Devices

Nicholas D. Lane, Sourav Bhattacharya, Petko Georgiev, Claudio Forlivesi, Fahim Kawsar
2015 Proceedings of the 2015 International Workshop on Internet of Things towards Applications - IoT-App '15  
The aim of this investigation is to begin to build knowledge of the performance characteristics, resource requirements and the execution bottlenecks for deep learning models when being used to recognize  ...  In this paper, we present the first -albeit preliminary -measurement study of common deep learning models (such as Convolutional Neural Networks and Deep Neural Networks) on representative mobile and embedded  ...  Moreover, many deep models remain out of reach due to their shear complexity; while AlexNet (60.9M parameters) barely executes on study hardware, the latest deep methods for important tasks are much larger  ... 
doi:10.1145/2820975.2820980 dblp:conf/sensys/LaneBGFK15 fatcat:m6omcl5synckzd7ukuvw77qaea

JALAD: Joint Accuracy-And Latency-Aware Deep Structure Decoupling for Edge-Cloud Execution

Hongshan Li, Chenghao Hu, Jingyan Jiang, Zhi Wang, Yonggang Wen, Wenwu Zhu
2018 2018 IEEE 24th International Conference on Parallel and Distributed Systems (ICPADS)  
Conventional cloud-based approaches usually run the deep models in data center servers, causing large latency because a significant amount of data has to be transferred from the edge of network to the  ...  A practical and critical problem thus has emerged: how to effectively deploy the deep neural network models such that they can be executed efficiently.  ...  Another solution is to run a mini-version of the original deep network models on edge devices, usually a compressed version of the original deep network model exported by pruning [5] - [7] or quantization  ... 
doi:10.1109/padsw.2018.8645013 dblp:conf/icpads/LiHJWWZ18 fatcat:m44nvmrc7neq5nl5lfmrzhvppe

DeepX: A Software Accelerator for Low-Power Deep Learning Inference on Mobile Devices

Nicholas D. Lane, Sourav Bhattacharya, Petko Georgiev, Claudio Forlivesi, Lei Jiao, Lorena Qendro, Fahim Kawsar
2016 2016 15th ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN)  
The foundation of DeepX is a pair of resource control algorithms, designed for the inference stage of deep learning, that: (1) decompose monolithic deep model network architectures into unit-blocks of  ...  In this work, we present the design and implementation of DeepX, a software accelerator for deep learning execution.  ...  The inputs to DAD include: (1) the deep model to be executed, (2) a set of performance goals (one or more metrics from: energy, execution time, model error).  ... 
doi:10.1109/ipsn.2016.7460664 dblp:conf/ipsn/LaneBGFJQK16 fatcat:asstog44mvfxtbc6icz7uma3nu

DEEP: A Provenance-Aware Executable Document System [chapter]

Huanjia Yang, Danius T. Michaelides, Chris Charlton, William J. Browne, Luc Moreau
2012 Lecture Notes in Computer Science  
We wish to thank Richard Parker and our other colleagues at the Centre for Multilevel Modelling for their input into the design of our system.  ...  Acknowledgments This researched was conducted as part of the E-Stat project, funded by the ESRC (RES-149-25-1084) under the Digital Social Research programme.  ...  Provenance data model for an executable document system On the basis of the requirements and the Deep document file structure discussed in the previous section, we present a provenance data model to describe  ... 
doi:10.1007/978-3-642-34222-6_3 fatcat:dweobjxrjjd4restrgrzy7anga

Comprehensive and Efficient Data Labeling via Adaptive Model Scheduling [article]

Mu Yuan, Lan Zhang, Xiang-Yang Li, Hui Xiong
2020 arXiv   pre-print
of these models to execute on each data item, aiming to maximize the value of the model output (e.g., the number of high-confidence labels).  ...  Numerous models were proposed to label various data and many approaches were designed to enhance the ability of deep learning models or accelerate them.  ...  Specifically, given a set of deep learning models and a stream of raw data to be processed, we propose to adaptively schedule model execution on each piece of input data to maximize the value of extracted  ... 
arXiv:2002.05520v1 fatcat:vnsdou2cofe4vpifrajknymk6y

Actor Critic Deep Reinforcement Learning for Neural Malware Control

Yu Wang, Jack Stokes, Mady Marinescu
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
This AC model dynamically predicts the best time to halt the file's execution based on a sequence of system API calls.  ...  Recent research uses a deep reinforcement learning (DRL) model employing a Deep Q-Network (DQN) to learn when to halt the emulation of a file.  ...  When the file's execution is controlled by a deep reinforcement learning model, one possibility is that the model does not terminate before reaching the end of the event sequence.  ... 
doi:10.1609/aaai.v34i01.5449 fatcat:dizfmtdvivglzlo2by7vaosqqq

Symbolic Techniques for Deep Learning: Challenges and Opportunities [article]

Belinda Fang, Elaine Yang, Fei Xie
2020 arXiv   pre-print
As the number of deep learning frameworks increase and certain ones gain popularity, it spurs the discussion of what methodologies are employed by these frameworks and the reasoning behind them.  ...  The goal of this survey is to study how symbolic techniques are utilized in deep learning.  ...  in the context of deep learning models.  ... 
arXiv:2010.02727v1 fatcat:zbvkwb3ewza7rmodjws4w6ysbu

Bring Your Own Codegen to Deep Learning Compiler [article]

Zhi Chen, Cody Hao Yu, Trevor Morris, Jorn Tuyls, Yi-Hsiang Lai, Jared Roesch, Elliott Delaye, Vin Sharma, Yida Wang
2021 arXiv   pre-print
Our framework provides users flexible and easy-to-use interfaces to partition their models into segments that can be executed on "the best" processors to take advantage of the powerful computation capability  ...  However, to achieve high model coverage with high performance, each accelerator vendor has to develop a full compiler stack to ingest, optimize, and execute the DNNs.  ...  (a) The process of loading a runtime module of the compiled model and executing the first operator on the host. (b) The process of executing an external function call on the accelerator.  ... 
arXiv:2105.03215v1 fatcat:4bjuy7tsdzfdpnkmki55q5jlim

HeNet: A Deep Learning Approach on Intel^ Processor Trace for Effective Exploit Detection [article]

Li Chen, Salmin Sultana, Ravi Sahita
2018 arXiv   pre-print
The low-level model is a per-application behavior model, trained via transfer learning on a time-series of images generated from control flow trace of an execution.  ...  Deep learning-based malware detection has so far focused on analyzing executable files and runtime API calls.  ...  On the other hand, it highlights the advantage of deep learning in greatly saving the cost of feature engineering. 2) Top-level ensemble model performance: The ensemble model aggregates the low level execution  ... 
arXiv:1801.02318v1 fatcat:5lnx2pj7sjggnbdmbliatqizla

DeepForge: A Scientific Gateway for Deep Learning

Brian Broll, Miklos Maroti, Peter Volgyesi, Akos Ledeczi
2018 Figshare  
We introduce DeepForge, a gateway to deep learning for the scientific community. DeepForge is designed to lower the barrier to entry and facilitate the rapid development of deep learning models.  ...  This represents an interdisciplinary approach to facilitating deep learning development as it leverages the strengths of Model Integrated Computing to provide a powerful hybrid textual-visual programming  ...  DeepForge supports the execution of machine learning pipelines on a distributed environment.  ... 
doi:10.6084/m9.figshare.7092272.v1 fatcat:2fbm6gji4vbzbctis7em6bzqti
« Previous Showing results 1 — 15 out of 418,985 results