Filters








919 Hits in 5.7 sec

Universal Distribution of Saliencies for Pruning in Layered Neural Networks

J. Gorodkin, L. K. Hansen, B. Lautrup, S. A. Solla
1997 International Journal of Neural Systems  
We focus on two-layer networks with either a linear or nonlinear output unit, and obtain analytic expressions for the distribution of saliencies and their logarithms.  ...  A better understanding of pruning methods based on a ranking of weights according to their saliency in a trained network requires further information on the statistical properties of such saliencies.  ...  This research was supported by the Danish Research Councils for the Natural and Technical Sciences through the Danish Computational Neural Network Center (CONNECT) and the Danish National Research Foundation  ... 
doi:10.1142/s0129065797000471 fatcat:47bxt6gvyjdsresbfg4nejhlby

Optimizing Artificial Neural Networks using Cat Swarm Optimization Algorithm

John Paul T. Yusiong
2012 International Journal of Intelligent Systems and Applications  
Experiments performed on benchmark datasets taken from the UCI machine learning repository show that the proposed CSONN-OBD is an effective tool for training neural networks.  ...  An Artificial Neural Network (ANN) is an abstract representation of the biological nervous system which has the ability to solve many complex problems.  ...  This work was supported by the University of the Philippines Visayas In-House Research Program under grant no. SP10-06.  ... 
doi:10.5815/ijisa.2013.01.07 fatcat:ywzdhunqijeb7lf6d56dnlbehm

Adaptive Dynamic Pruning for Non-IID Federated Learning [article]

Sixing Yu, Phuong Nguyen, Ali Anwar, Ali Jannesari
2021 arXiv   pre-print
However, the limited computing power and energy constraints of edge devices hinder the adoption of FL for both model training and deployment, especially for the resource-hungry Deep Neural Networks (DNNs  ...  In this paper, we present an adaptive pruning scheme for edge devices in an FL system, which applies dataset-aware dynamic pruning for inference acceleration on Non-IID datasets.  ...  The key idea of network pruning is to permanently remove deep neural networks' (DNNs) redundant weights by evaluating the saliency of neurons for the input data (Gao et al., 2019) .  ... 
arXiv:2106.06921v1 fatcat:4igg3t2h4ff45fjs5m4hmc5jei

Network Compression for Machine-Learnt Fluid Simulations [article]

Peetak Mitra, Vaidehi Venkatesan, Nomit Jangid, Ashwati Nambiar, Dhananjay Kumar, Vignesh Roa, Niccolo Dal Santo, Majid Haghshenas, Shounak Mitra, David Schmidt
2021 arXiv   pre-print
In this study, we explore the applicability of pruning and quantization (FP32 to int8) methods for one such application relevant to modeling fluid turbulence.  ...  For full physics emulators, the cost of network inference is often trivial. However, in the current paradigm of data-driven fluid mechanics models are built as surrogates for complex sub-processes.  ...  Visualizing the distribution of network weights in a low-dimensional space for the same layer, using t-SNE [Van der Maaten & Hinton (2008)], shows a similar trend of overlapping regions of similarity,  ... 
arXiv:2103.00754v1 fatcat:blvmlfd6azgfbdzwi5jepmrloq

Pruning of recurrent neural models: an optimal brain damage approach

Patryk Chaber, Maciej Ławryńczuk
2018 Nonlinear dynamics  
This paper considers the problem of pruning recurrent neural models of perceptron type with one hidden layer which may be used for modelling of dynamic system.  ...  In order to reduce the number of model parameters (i.e. the number of weights), the Optimal Brain Damage (OBD) pruning algorithm is adopted for the recurrent neural models.  ...  The above formulae are universal for the considered recurrent neural model.  ... 
doi:10.1007/s11071-018-4089-1 fatcat:anvoha6tfbga5oq4cmewz7ziee

Global Biased Pruning Considering Layer Contribution

Zheng Huang, Li Li, Hailin Sun
2020 IEEE Access  
Most existing methods for filter pruning only consider the role of the filter itself, ignoring the characteristics of the layer.  ...  Convolutional neural networks (CNNs) have made impressive achievements in many areas, but these successes are limited by storage and computing costs.  ...  INDEX TERMS deep learning, network pruning, convolutional neural networks I.  ... 
doi:10.1109/access.2020.3025130 fatcat:jb23kba2nbbglhrvss5rigrgkq

Pruning Algorithms to Accelerate Convolutional Neural Networks for Edge Applications: A Survey [article]

Jiayi Liu, Samarth Tripathi, Unmesh Kurup, Mohak Shah
2020 arXiv   pre-print
With the general trend of increasing Convolutional Neural Network (CNN) model sizes, model compression and acceleration techniques have become critical for the deployment of these models on edge devices  ...  The survey covers the overarching motivation for pruning, different strategies and criteria, their advantages and drawbacks, along with a compilation of major pruning techniques.  ...  [53] used the scaling factor of the Batch Normalization layer as the saliency measure.  ... 
arXiv:2005.04275v1 fatcat:2w4d65rebjbvvpeddx6ygc6hge

Learning compact ConvNets through filter pruning based on the saliency of a feature map

Zhoufeng Liu, Xiaohui Liu, Chunlei Li, Shumin Ding, Liang Liao
2021 IET Image Processing  
Among the methods mentioned in various literature, filter pruning is a crucial method for constructing lightweight networks.  ...  With the performance increase of convolutional neural network (CNN), the disadvantages of CNN's high storage and high power consumption are followed.  ...  Direction Team in Zhongyuan University of Technology.  ... 
doi:10.1049/ipr2.12338 fatcat:75u375ziqfgtbjjw4ifuoxlgdy

System Identification With General Dynamic Neural Networks And Network Pruning

Christian Endisch, Christoph Hackl, Dierk Schröder
2008 Zenodo  
This paper presents an exact pruning algorithm with adaptive pruning interval for general dynamic neural networks (GDNN). GDNNs are artificial neural networks with internal dynamics.  ...  During parameter optimization with the Levenberg- Marquardt (LM) algorithm irrelevant weights of the dynamic neural network are deleted in order to find a model for the plant as simple as possible.  ...  CONCLUSION In this paper it is shown, that network pruning not only works in static neural networks but can also be applied to dynamic neural networks.  ... 
doi:10.5281/zenodo.1080759 fatcat:36tawnkq45dqxddwvduxjkpwqy

Overcoming Long-term Catastrophic Forgetting through Adversarial Neural Pruning and Synaptic Consolidation [article]

Jian Peng, Bo Tang, Hao Jiang, Zhuo Li, Yinjie Lei, Tao Lin, Haifeng Li
2021 IEEE Transactions on Neural Networks and Learning Systems   accepted
Artificial neural networks face the well-known problem of catastrophic forgetting.  ...  Inspired by the memory consolidation mechanism in mammalian brains with synaptic plasticity, we propose a confrontation mechanism in which Adversarial Neural Pruning and synaptic Consolidation (ANPyC)  ...  a neural network to sequentially learn multiple tasks is of great significance for expanding the applicability of neural networks in realistic human application scenarios.  ... 
doi:10.1109/tnnls.2021.3056201 pmid:33577459 arXiv:1912.09091v2 fatcat:glic2itroraa7jpicjaamjljsu

Differentiable Network Pruning for Microcontrollers [article]

Edgar Liberis, Nicholas D. Lane
2021 arXiv   pre-print
In this work, we present a differentiable structured network pruning method for convolutional neural networks, which integrates a model's MCU-specific resource usage and parameter importance feedback to  ...  Orders of magnitude less storage, memory and computational capacity, compared to what is typically required to execute neural networks, impose strict structural constraints on the network architecture  ...  SNIP: Single-shot is the state of neural network pruning?  ... 
arXiv:2110.08350v2 fatcat:7jmlafbgzjexbbtdyslmlx6ig4

Methodological Challenges in Neural Spatial Interaction Modelling: The Issue of Model Selection [chapter]

Manfred M. Fischer
2000 Advances in Spatial Science  
Rigorous mathematical proofs for the universality of such feedforward neural network models (see, among others, Hornik, Stinchcombe and White 1989 ) establish the neural spatial interaction models as  ...  First, a summarized description of single hidden layer neural spatial interaction is given in the next section.  ... 
doi:10.1007/978-3-642-59787-9_6 fatcat:4aajvft76ff33algaho7iawo6m

Compression of Neural Machine Translation Models via Pruning

Abigail See, Minh-Thang Luong, Christopher D. Manning
2016 Proceedings of The 20th SIGNLL Conference on Computational Natural Language Learning  
for the different classes of weights in the NMT architecture.  ...  We demonstrate the efficacy of weight pruning as a compression technique for a state-of-the-art NMT system.  ...  Lastly, we acknowledge NVIDIA Corporation for the donation of Tesla K40 GPUs.  ... 
doi:10.18653/v1/k16-1029 dblp:conf/conll/SeeLM16 fatcat:iaexhxacdravvisc34lzqsygve

FreezeNet: Full Performance by Reduced Storage Costs [article]

Paul Wimmer, Jens Mehnert, Alexandru Condurache
2020 arXiv   pre-print
Pruning generates sparse networks by setting parameters to zero.  ...  In our experiments we show that FreezeNets achieve good results, especially for extreme freezing rates.  ...  Backpropagation in Neural Networks To simplify the backpropagation formulas, we will deal with a feed-forward, fully connected neural network. Similar equations hold for convolutional layers [21] .  ... 
arXiv:2011.14087v1 fatcat:mazpvgnxxnaw7psj2gev7ovhde

Compression of Neural Machine Translation Models via Pruning [article]

Abigail See, Minh-Thang Luong, Christopher D. Manning
2016 arXiv   pre-print
for the different classes of weights in the NMT architecture.  ...  We demonstrate the efficacy of weight pruning as a compression technique for a state-of-the-art NMT system.  ...  Lastly, we acknowledge NVIDIA Corporation for the donation of Tesla K40 GPUs.  ... 
arXiv:1606.09274v1 fatcat:urda6y32wbbcjoii4ugpftvbwa
« Previous Showing results 1 — 15 out of 919 results