91,840 Hits in 2.8 sec

Algorithms for Verifying Deep Neural Networks [article]

Changliu Liu, Tomer Arnon, Christopher Lazarus, Clark Barrett, Mykel J. Kochenderfer
2020 arXiv   pre-print
Deep neural networks are widely used for nonlinear function approximation with applications ranging from computer vision to control.  ...  Although these networks involve the composition of simple arithmetic operations, it can be very challenging to verify whether a particular network satisfies certain input-output properties.  ...  The authors would like to thank many of the authors of the referenced papers for their help in clarifying their algorithms and reviewing early drafts of this survey: Weiming Xiang, Taylor Johnson, Hoang-Dung  ... 
arXiv:1903.06758v2 fatcat:25pqxtxpfzfz7phnnsx53q3j5y

Algorithms for Verifying Deep Neural Networks

Changliu Liu, Tomer Arnon, Christopher Lazarus, Christopher Strong, Clark Barrett, Mykel J. Kochenderfer
2020 Foundations and Trends® in Optimization  
Foundations and Trends ® in Optimization publishes survey and tutorial articles in the following topics: • algorithm design, analysis, and implementation (especially, on modern computing platforms • models  ...  economics and finance, engineering design, scheduling and resource allocation, and other areas Information for Librarians  ...  This article surveys a class of methods that are capable of formally verifying properties of deep neural networks over the full input space.  ... 
doi:10.1561/2400000035 fatcat:udnpbqyaqbeatcjrbkohospau4

Evaluation of Activity Monitoring Algorithm based on Smart Approaches

Vivekanadam B
2020 Journal of Electronics and Informatics  
The paper evaluates certain artificial intelligence based deep learning techniques for finding a suitable approach for monitoring the listener's activity in real time.  ...  To improve the quality of such online classes, it is mandatory to verify the listener's activity.  ...  The convolution neural networks are the basic algorithms in the deep neural networks widely used for classifying the images.  ... 
doi:10.36548/jei.2020.3.004 fatcat:c7dvcjftgrgpdbd7vod4mgqbbu

Deep Binary Reinforcement Learning for Scalable Verification [article]

Christopher Lazarus, Mykel J. Kochenderfer
2022 arXiv   pre-print
We present an RL algorithm tailored specifically for BNNs. After training BNNs for the Atari environments, we verify robustness properties.  ...  The generalization power of neural networks combined with advances in RL algorithms has reignited the field of artificial intelligence.  ...  Potential avenues for future work include investigating how binarized neural networks can be used in more sophisticated versions of value function-based RL algorithms such as the Double Deep Q-Network  ... 
arXiv:2203.05704v1 fatcat:a4cyawb24jfa3fu6hlwzsn6fiu

Recognition of Combat Intention with Insufficient Expert Knowledge

Wang-wang ZHOU, Jie-yong ZHANG, Nan-nan GU, Guo-qiang YAN
2018 DEStech Transactions on Computer Science and Engineering  
neural network is designed.  ...  By introducing the ReLU activation function and the Adam optimization algorithm, the convergence speed of the model is improved, and the local optimization is effectively prevented.  ...  BP neural network model, the Adam+Sigmiod deep neural network model and the Adam+RuLU deep neural network model.  ... 
doi:10.12783/dtcse/cmsam2018/26561 fatcat:qxcqivwbwfcb3g7e5zwktg472q

A hypothesize-and-verify framework for Text Recognition using Deep Recurrent Neural Networks [article]

Anupama Ray, Sai Rajeswar, Santanu Chaudhury
2015 arXiv   pre-print
deep neural network.  ...  Deep LSTM is an ideal candidate for text recognition.  ...  2014, pp. 4823–4827. network and verifies the performance of the deep neural [7] M.  ... 
arXiv:1502.07540v1 fatcat:lmptvd5wxfcmhddkhh5bkn44ue

An Image Classification Algorithm Based on Multidomain Convolution Neural Network

Tao-wei JIANG, Meng-yu ZHU, Yong-hao HAI, De-zhuang KONG
2018 DEStech Transactions on Computer Science and Engineering  
Deep Convolutional Neural Networks (CNNs) have outperformed humans in many computer vision tasks, such as object recognition and image classification, but it is almost impossible to run a large-scale CNN  ...  For the limited computing platform and application scenarios, we propose a novel CNN architecture: Multi-Domain CNN (MD-CNN).  ...  With the development of deep artificial neural network technology, deep neural network model based on CNN has made great progress in the image classification task.  ... 
doi:10.12783/dtcse/wcne2017/19883 fatcat:anl5anbmwvhhvojphzvmy5xy6i

A Deep Learning Based Transmission Algorithm for Mobile Device-to-Device Networks

Tae-Won Ban, Woongsup Lee
2019 Electronics  
D2D networks based on deep learning with a convolutional neural network (CNN).  ...  In order to train the CNN and verify the trained CNN, we obtain data samples from a suboptimal algorithm.  ...  feasibility of deep learning has been verified in D2D networks.  ... 
doi:10.3390/electronics8111361 fatcat:iocdf77snraljnkkcrntw3pjdm

Using Pretrained AlexNet Deep Learning Neural Network for Recognition of Underwater Objects
Uporaba pretreniranih AlexNet neuronskih mreža dubokog učenja za prepoznavanje podvodnih objekata

Piotr Szymak, Institute of Electrical Enginnering and Automatics Polish Naval Academy Gdynia Poland, Marek Gasiorowski
2020 Naše More (Dubrovnik)  
Based on several examples included in the literature, the object recognition algorithm proposed in the paper is based on the deep neural network.  ...  In the research, the network and training algorithms accessible in the Matlab have been used.  ...  partners: Polish Naval Academy AMW -the leader, Cracow University of Technology PK, Industrial Institute of Automatics and Measurement PIAP, Forkos Company, German partners: Bundeswehr Technical Center for  ... 
doi:10.17818/nm/2020/1.2 fatcat:s5xp6fvpjjbuxc7v2supcrooke

Research on Prediction of Ground Settlement of Deep Foundation Pit Based on Improved PSO-BP Neural Network

Liu Yuhao, Feng Xiao, H. Zhang, H. Abdul Aziz
2021 E3S Web of Conferences  
In view of the limitations of the existing prediction methods for ground subsidence of deep foundation pit, a BP neural network prediction model based on improved particle swarm optimization algorithm  ...  weight and threshold of the BP neural network.  ...  used BP neural network to predict the deformation of deep foundation pit.  ... 
doi:10.1051/e3sconf/202127601014 fatcat:a2xavfo45rcarc574mageamvm4

Towards Sample Efficient Agents through Algorithmic Alignment [article]

Mingxuan Li, Michael L. Littman
2021 arXiv   pre-print
We believe this would open up a new avenue for structured agent design. See for the code.  ...  The main idea is that the agent should be guided by structured non-neural-network algorithms like dynamic programming.  ...  Deep Graph Value Network Due to the space limitation, for general context of MDP and graph neural network, please refer to Sutton and Barto (2018) ; Battaglia, Hamrick et al. (2018) .  ... 
arXiv:2008.03229v5 fatcat:wppzowq7j5hddmg5uqyooisyfq

Toward Scalable Verification for Safety-Critical Deep Networks [article]

Lindsey Kuper, Guy Katz, Justin Gottschlich, Kyle Julian, Clark Barrett, Mykel Kochenderfer
2018 arXiv   pre-print
The increasing use of deep neural networks for safety-critical applications, such as autonomous driving and flight control, raises concerns about their safety and reliability.  ...  Formal verification can address these concerns by guaranteeing that a deep learning system operates as intended, but the state of the art is limited to small systems.  ...  INTRODUCTION Machine learning systems, and, in particular, deep neural networks (DNNs), are becoming a widely used and effective means for tackling complex, real-world problems [4] .  ... 
arXiv:1801.05950v2 fatcat:up5f4gtbrfe3hktgeuk2vzfomm

Delta Ruled Fully Recurrent Deep Learning for Finger-Vein Verification

However, the verification accuracy of existing algorithms was not sufficient. Also, the amount of time required for verifying the input finger vein image was more.  ...  In order to overcome such limitations, Delta Ruled Fully Recurrent Deep Learning (DRFRDL) technique is proposed.  ...  in a fully recurrent deep neural network structure.  ... 
doi:10.35940/ijitee.b7303.129219 fatcat:smnqt2mtdza3dlcgcfrnvmesxa

Deep Anomaly Detection via Morphological Transformations

Taehyeon Kim, Yoonsik Choe
2020 Proceedings (MDPI)  
The goal of deep anomaly detection is to identify abnormal data by utilizing a deep neural network trained by a normal training dataset.  ...  Additionally, we present a kernel size loss to enhance the proposed neural networks' morphological feature representation power.  ...  Introduction Deep anomaly detection means verifying abnormal data via a deep neural network trained by normal instances.  ... 
doi:10.3390/asec2020-07887 fatcat:dx2q42s7pngu5lzbxsyhomafeu

Facemask Detection Algorithm on COVID Community Spread Control using EfficientNet Algorithm

Vivekanadam Balasubramaniam
2021 Journal of Soft Computing Paradigm  
Recently, deep learning algorithms are emerging as a fast growing application, which has been developed for performing huge number of analysis and detection process.  ...  Henceforth, this paper proposes a deep learning based facemask detection process for automating the human effort involved in monitoring process.  ...  neural network.  ... 
doi:10.36548/jscp.2021.2.005 fatcat:zfpiz4zbmzevpp6s2s2bmcj6vq
« Previous Showing results 1 — 15 out of 91,840 results