Filters








6,017 Hits in 3.1 sec

Multiple attractors, long chaotic transients, and failure in small-world networks of excitable neurons

Hermann Riecke, Alex Roxin, Santiago Madruga, Sara A. Solla
2007 Chaos  
These dynamics depend in large part on the fraction of long-range connections or 'short-cuts' and the delay in the neuronal interactions.  ...  We show how this functional form arises in the ensemble-averaged activity if each network realization has a characteristic failure-time which is exponentially distributed.  ...  Other modifications to the distribution of shortcut lengths might include allowing for a nonuniform distribution.  ... 
doi:10.1063/1.2743611 pmid:17614697 fatcat:p7hseoy3uvbiflo7gkcjzrnthm

Deep Learning Made Easier by Linear Transformations in Perceptrons

Tapani Raiko, Harri Valpola, Yann LeCun
2012 Journal of machine learning research  
We transform the outputs of each hidden neuron in a multi-layer perceptron network to have zero output and zero slope on average, and use separate shortcut connections to model the linear dependencies  ...  The experiments include both classification of small images and learning a lowdimensional representation for images by using a deep unsupervised auto-encoder network.  ...  Biases are drawn from a uniform distribution between ±0.5. Weights for all shortcut connections are initialized to zero. Learning Rate We use a hand-set learning rate γ i for each problem.  ... 
dblp:journals/jmlr/RaikoVL12 fatcat:xxsbo33og5enbotndqjb64l564

DDU-Nets: Distributed Dense Model for 3D MRI Brain Tumor Segmentation [article]

Hanxiao Zhang, Jingxiong Li, Mali Shen, Yaqi Wang, Guang-Zhong Yang
2020 arXiv   pre-print
In this paper, three patterns (cross-skip, skip-1 and skip-2) of distributed dense connections (DDCs) are proposed to enhance feature reuse and propagation of CNNs by constructing tunnels between key layers  ...  For better detecting and segmenting brain tumors from multi-modal 3D MR images, CNN-based models embedded with DDCs (DDU-Nets) are trained efficiently from pixel to pixel with a limited number of parameters  ...  We also denied the opening operation solution used to denoise for the postprocessing which is replaced by connected component processing.  ... 
arXiv:2003.01337v1 fatcat:nux66zsz2ngghmkgp5bgaxgvta

Low-Congestion Shortcuts without Embedding

Bernhard Haeupler, Taisuke Izumi, Goran Zuzic
2016 Proceedings of the 2016 ACM Symposium on Principles of Distributed Computing - PODC '16  
Distributed optimization algorithms are frequently faced with solving sub-problems on disjoint connected parts of a network.  ...  This directly leads to fast O(D log^O(1) n) distributed algorithms for MST and Min-Cut approximation, given that one can efficiently construct these shortcuts in a distributed manner.  ...  If such a connected component intersects P i we call it a block component. Furthermore, we define the block parameter b of H to be any upper bound to the number of block components for all parts.  ... 
doi:10.1145/2933057.2933112 dblp:conf/podc/HaeuplerIZ16 fatcat:yyzfj4yzoja35iyd2cpw4yela4

Learning Deep Representations Using Convolutional Auto-encoders with Symmetric Skip Connections [article]

Jianfeng Dong, Xiao-Jiao Mao, Chunhua Shen, Yu-Bin Yang
2017 arXiv   pre-print
We empirically show that symmetric shortcut connections are very important for learning abstract representations via image reconstruction.  ...  The architecture we use is a convolutional auto-encoder network with symmetric shortcut connections.  ...  While in our method, the shortcut connections pass feature maps to decoder and also help to back-propagate gradients during training.  ... 
arXiv:1611.09119v2 fatcat:uijwlfho3bc2zm2c7nje23mlx4

Thrifty Label Propagation: Fast Connected Components for Skewed-Degree Graphs

Mohsen Koohi Esfahani, Peter Kilpatrick, Hans Vandierendonck
2021 2021 IEEE International Conference on Cluster Computing (CLUSTER)  
We investigate the implications of the skewed degree distribution of real-world graphs on their connectivity and we use these features to introduce Thrifty Label Propagation as a structureaware CC algorithm  ...  obtained by incorporating 4 fundamental optimization techniques in the Label Propagation CC algorithm.  ...  ACKNOWLEDGEMENTS We are grateful for the constructive feedback of our CLUS-TER reviewers.  ... 
doi:10.1109/cluster48925.2021.00042 fatcat:tuaugzgh5fgh5bm2iwg5hzuyz4

scDeepC3: scRNA-seq Deep Clustering by A Skip AutoEncoder Network with Clustering Consistency [article]

Gang Wu, Junjun Jiang, Xianming Liu
2022 bioRxiv   pre-print
In particular, we introduce an adaptive shortcut connection layer to directly add the shallow-layer (encoder) features to deep-layer (decoder).  ...  Based on the above observation, for the scRNA-seq clustering analysis, we design the adaptive shortcut connection layer and extend standard AutoEncoder structure with shortcut connection layer.  ...  Because the training process for the deep clustering model is unsupervised and we have no label information for evaluation.  ... 
doi:10.1101/2022.06.05.494891 fatcat:r3a6ogaez5drnbmdal4ukji5xe

Unified DeepLabV3+ for Semi-Dark Image Semantic Segmentation

Mehak Maqbool Memon, Manzoor Ahmed Hashmani, Aisha Zahid Junejo, Syed Sajjad Rizvi, Kamran Raza
2022 Sensors  
Extensive experimental analysis performed over a CamVid dataset confirmed the applicability of the proposed solution for autonomous vehicles and robotics for outdoor settings.  ...  Semantic segmentation for accurate visual perception is a critical task in computer vision.  ...  For any DCNN, the objective function is given as Equation (3). where 𝜃 is the parameter vector for DCNN. The pixel label distributions are calculated using Equation (4).  ... 
doi:10.3390/s22145312 pmid:35890992 pmcid:PMC9324997 fatcat:jqmwgcqs4vfspilnqn2hvy7r2m

Low-Congestion Shortcut and Graph Parameters

Naoki Kitamura, Hirotaka Kitagawa, Yota Otachi, Taisuke Izumi, Michael Wagner
2019 International Symposium on Distributed Computing  
Distributed graph algorithms in the standard CONGEST model often exhibit the time-complexity lower bound ofΩ( √ n + D) rounds for many global problems, where n is the number of nodes and D is the diameter  ...  [Distributed Computing 2006]. (3) We show that bounding clique-width does not help the construction of good shortcuts by presenting a network topology of clique-width six where the construction of MST  ...  We conclude this paper posing three related open problems. (1) Can we have good shortcuts for D ≥ 5? (2) Can we have good shortcuts for k-clique width where k ≤ 5?  ... 
doi:10.4230/lipics.disc.2019.25 dblp:conf/wdag/KitamuraKOI19 fatcat:az5dhls7vnbkxb34xtyq5pk2dq

Simple Concurrent Connected Components Algorithms

S. Cliff Liu, Robert E. Tarjan
2022 ACM Transactions on Parallel Computing  
We study a class of simple algorithms for concurrently computing the connected components of an n -vertex, m -edge graph.  ...  Our results show that even a basic problem like connected components still has secrets to reveal.  ...  We thank Pei-Duo Yu for discovering that Lemma 6 in [17] does not hold for P and E, invalidating our proof of Theorem 13 in [17] for these algorithms.  ... 
doi:10.1145/3543546 fatcat:z72s5zghqjbtthpedwp3cfoowe

Feature Fusion Based on Convolutional Neural Network for SAR ATR

Shi-Qi Chen, Rong-Hui Zhan, Jie-Min Hu, Jun Zhang, L. Long, Y. Li, X. Li, Y. Dai, H. Yang
2017 ITM Web of Conferences  
Inspired by the more efficient distributed training such as inception architecture and residual network, a new feature fusion structure which jointly exploits all the merits of each version is proposed  ...  to the target label among T classes. The prediction ( ) i j y of jth class for ith input is passed through softmax function to yield a normalized probability distribution over classes.  ...  As for the second pattern of feature fusion, an asymmetric shortcut connection block is inserted into the network between each convolution layer and pooling layer, with different asymmetric convolution  ... 
doi:10.1051/itmconf/20171205001 fatcat:3dgx3yrtg5fjzaeyonoisvvz2q

Exact solution of site and bond percolation on small-world networks

Cristopher Moore, M. E. J. Newman
2000 Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics  
We study percolation on small-world networks, which has been proposed as a simple model of the propagation of disease.  ...  governing this transition, and the mean and variance of the distribution of cluster sizes (disease outbreaks) below the transition.  ...  for useful comments.  ... 
doi:10.1103/physreve.62.7059 pmid:11102061 fatcat:rce6ilpvrvd3nlgeevfwugwiai

Self-Stabilizing Supervised Publish-Subscribe Systems [article]

Michael Feldmann, Christina Kolb, Christian Scheideler, Thim Strothmann
2017 arXiv   pre-print
That is, in addition to stabilizing the overlay network, every subscriber of a topic will eventually know all of the publications that have been issued so far for that topic.  ...  subscribe or unsubscribe operation is just a constant in a legitimate state, and the communication work of checking whether the system is still in a legitimate state is just a constant on expectation for  ...  We now sketch why eventually all subscribers in a connected component C get their correct label.  ... 
arXiv:1710.08128v2 fatcat:566n7pbpdbcnxlbbgqhv2sesmu

Unsupervised anomaly detection for a Smart Autonomous Robotic Assistant Surgeon (SARAS)using a deep residual autoencoder [article]

Dinesh Jackson Samuel, Fabio Cuzzolin
2021 arXiv   pre-print
The idea is to make the autoencoder learn the 'normal' distribution of the data and detect abnormal events deviating from this distribution by measuring the reconstruction error.  ...  In this work we thus propose an unsupervised approach to anomaly detection for robotic-assisted surgery based on deep residual autoencoders.  ...  When compared with the concatenated shortcuts in U-Net [23] , these residual connections have the property of minimising the number of training parameters of the model and of enhancing learning by propagating  ... 
arXiv:2104.11008v1 fatcat:bvpmpb7gjjfzzg35dixagf7b2m

SOON: A Scalable Self-organized Overlay Network for Distributed Information Retrieval [chapter]

Juan Li, Son Vuong
2008 Lecture Notes in Computer Science  
For this purpose, we have designed an algorithm to extract a node's ontology summary and use that summary to compute the semantic similarity between nodes.  ...  However, the distributed, heterogeneous, and unstructured nature of the system makes this issue very challenging.  ...  Unlike ShortCut and random-walk approaches, which only create query propagating overhead, SOON also creates overhead for maintaining neighborhood relationship.  ... 
doi:10.1007/978-3-540-87353-2_1 fatcat:gc5o7mouwfbhfmkidxrchfd56y
« Previous Showing results 1 — 15 out of 6,017 results