000218985 001__ 218985
000218985 005__ 20180913063728.0
000218985 037__ $$aCONF
000218985 245__ $$aConvolutional Neural Networks on Graphs with Fast Localized Spectral Filtering
000218985 269__ $$a2016
000218985 260__ $$c2016
000218985 336__ $$aConference Papers
000218985 520__ $$aIn this work, we are interested in generalizing convolutional neural networks (CNNs) from low-dimensional regular grids, where image, video and speech are represented, to high-dimensional irregular domains, such as social networks, brain connectomes or words' embedding, represented by graphs. We present a formulation of CNNs in the context of spectral graph theory, which provides the necessary mathematical background and efficient numerical schemes to design fast localized convolutional filters on graphs. Importantly, the proposed technique offers the same linear computational complexity and constant learning complexity as classical CNNs, while being universal to any graph structure. Experiments on MNIST and 20NEWS demonstrate the ability of this novel deep learning system to learn local, stationary, and compositional features on graphs.
000218985 6531_ $$agraphs
000218985 6531_ $$aconvolutional neural networks
000218985 6531_ $$adeep learning
000218985 6531_ $$agraph signal processing
000218985 700__ $$0249515$$aDefferrard, Michaël$$g226056
000218985 700__ $$0241065$$aBresson, Xavier$$g140163
000218985 700__ $$0240428$$aVandergheynst, Pierre$$g120906
000218985 7112_ $$aAdvances in Neural Information Processing Systems 29$$cBarcelona, Spain$$dDecember 5-10, 2016
000218985 8564_ $$uhttps://arxiv.org/abs/1606.09375$$zURL
000218985 8564_ $$uhttp://papers.nips.cc/paper/6081-convolutional-neural-networks-on-graphs-with-fast-localized-spectral-filtering$$zURL
000218985 8564_ $$uhttps://github.com/mdeff/cnn_graph$$zURL
000218985 909C0 $$0252392$$pLTS2$$xU10380
000218985 909CO $$ooai:infoscience.tind.io:218985$$pconf$$pSTI
000218985 917Z8 $$x226056
000218985 917Z8 $$x252028
000218985 917Z8 $$x144315
000218985 917Z8 $$x226056
000218985 937__ $$aEPFL-CONF-218985
000218985 973__ $$aEPFL$$rREVIEWED$$sPUBLISHED
000218985 980__ $$aCONF