Scale-Invariant Convolutional Neural Networks. Xu, Y., Xiao, T., Zhang, J., Yang, K., & Zhang, Z. *ArXiv e-prints*, November, 2014. Paper bibtex @article{Xu:2014wr,
author = {Xu, Yichong and Xiao, Tianjun and Zhang, Jiaxing and Yang, Kuiyuan and Zhang, Zheng},
title = {{Scale-Invariant Convolutional Neural Networks}},
journal = {ArXiv e-prints},
year = {2014},
volume = {cs.CV},
month = nov,
annote = {similar to "A. Kanazawa, A. Sharma, and D. Jacobs, {\textquotedblleft}Locally Scale-Invariant Convolutional Neural Networks,{\textquotedblright} arXiv, vol. cs.CV, Dec. 2014"
except that it's doing scaling on filters, not images.
also, in the end, concat is used, instead of max.
the derivation of this work is principled, see Eq. 2. Basically, an analogy between Eq. 2 and the equation for translation invariance can be made.
main problem of such work, is that they really don't work that well on standard datasets. See Section 3.4.
the error goes from 11.68-11.35, by using such a technique. it's really not that impressive.},
read = {Yes},
rating = {3},
date-added = {2017-05-05T17:35:44GMT},
date-modified = {2017-05-05T21:13:51GMT},
url = {http://arxiv.org/abs/1411.6369},
local-url = {file://localhost/Users/yimengzh/Documents/Papers3_revised/Library.papers3/Articles/2014/Xu/arXiv%202014%20Xu.pdf},
file = {{arXiv 2014 Xu.pdf:/Users/yimengzh/Documents/Papers3_revised/Library.papers3/Articles/2014/Xu/arXiv 2014 Xu.pdf:application/pdf}},
uri = {\url{papers3://publication/uuid/F597FE1C-2C8F-40EF-AFDE-A58CD59F9375}}
}

Downloads: 0

{"_id":"9pNq5YmwbziwhxDSS","bibbaseid":"xu-xiao-zhang-yang-zhang-scaleinvariantconvolutionalneuralnetworks-2014","downloads":0,"creationDate":"2017-06-08T12:53:07.197Z","title":"Scale-Invariant Convolutional Neural Networks","author_short":["Xu, Y.","Xiao, T.","Zhang, J.","Yang, K.","Zhang, Z."],"year":2014,"bibtype":"article","biburl":"https://leelabcnbc.github.io/lab-wiki/reference_library/machine_learning/neural_networks/bib.bib","bibdata":{"bibtype":"article","type":"article","author":[{"propositions":[],"lastnames":["Xu"],"firstnames":["Yichong"],"suffixes":[]},{"propositions":[],"lastnames":["Xiao"],"firstnames":["Tianjun"],"suffixes":[]},{"propositions":[],"lastnames":["Zhang"],"firstnames":["Jiaxing"],"suffixes":[]},{"propositions":[],"lastnames":["Yang"],"firstnames":["Kuiyuan"],"suffixes":[]},{"propositions":[],"lastnames":["Zhang"],"firstnames":["Zheng"],"suffixes":[]}],"title":"Scale-Invariant Convolutional Neural Networks","journal":"ArXiv e-prints","year":"2014","volume":"cs.CV","month":"November","annote":"similar to \"A. Kanazawa, A. Sharma, and D. Jacobs, “Locally Scale-Invariant Convolutional Neural Networks,” arXiv, vol. cs.CV, Dec. 2014\" except that it's doing scali\u0010ng on filters, not images. also, in the end, concat is used, instead of max. the derivation of this work is principled, see Eq. 2. Basically, an analogy between Eq. 2 and the equation for translation invariance can be made. main problem of such work, is that they really don't work that well on standard datasets. See Section 3.4. the error goes from 11.68-11.35, by using such a technique. it's really not that impressive.","read":"Yes","rating":"3","date-added":"2017-05-05T17:35:44GMT","date-modified":"2017-05-05T21:13:51GMT","url":"http://arxiv.org/abs/1411.6369","local-url":"file://localhost/Users/yimengzh/Documents/Papers3_revised/Library.papers3/Articles/2014/Xu/arXiv%202014%20Xu.pdf","file":"arXiv 2014 Xu.pdf:/Users/yimengzh/Documents/Papers3_revised/Library.papers3/Articles/2014/Xu/arXiv 2014 Xu.pdf:application/pdf","uri":"˘rlpapers3://publication/uuid/F597FE1C-2C8F-40EF-AFDE-A58CD59F9375","bibtex":"@article{Xu:2014wr,\nauthor = {Xu, Yichong and Xiao, Tianjun and Zhang, Jiaxing and Yang, Kuiyuan and Zhang, Zheng},\ntitle = {{Scale-Invariant Convolutional Neural Networks}},\njournal = {ArXiv e-prints},\nyear = {2014},\nvolume = {cs.CV},\nmonth = nov,\nannote = {similar to \"A. Kanazawa, A. Sharma, and D. Jacobs, {\\textquotedblleft}Locally Scale-Invariant Convolutional Neural Networks,{\\textquotedblright} arXiv, vol. cs.CV, Dec. 2014\"\n\nexcept that it's doing scali\u0010ng on filters, not images.\n\nalso, in the end, concat is used, instead of max.\n\n\nthe derivation of this work is principled, see Eq. 2. Basically, an analogy between Eq. 2 and the equation for translation invariance can be made.\n\n\nmain problem of such work, is that they really don't work that well on standard datasets. See Section 3.4.\n\nthe error goes from 11.68-11.35, by using such a technique. it's really not that impressive.},\nread = {Yes},\nrating = {3},\ndate-added = {2017-05-05T17:35:44GMT},\ndate-modified = {2017-05-05T21:13:51GMT},\nurl = {http://arxiv.org/abs/1411.6369},\nlocal-url = {file://localhost/Users/yimengzh/Documents/Papers3_revised/Library.papers3/Articles/2014/Xu/arXiv%202014%20Xu.pdf},\nfile = {{arXiv 2014 Xu.pdf:/Users/yimengzh/Documents/Papers3_revised/Library.papers3/Articles/2014/Xu/arXiv 2014 Xu.pdf:application/pdf}},\nuri = {\\url{papers3://publication/uuid/F597FE1C-2C8F-40EF-AFDE-A58CD59F9375}}\n}\n\n\n\n","author_short":["Xu, Y.","Xiao, T.","Zhang, J.","Yang, K.","Zhang, Z."],"key":"Xu:2014wr","id":"Xu:2014wr","bibbaseid":"xu-xiao-zhang-yang-zhang-scaleinvariantconvolutionalneuralnetworks-2014","role":"author","urls":{"Paper":"http://arxiv.org/abs/1411.6369"},"metadata":{"authorlinks":{}}},"search_terms":["scale","invariant","convolutional","neural","networks","xu","xiao","zhang","yang","zhang"],"keywords":[],"authorIDs":[],"dataSources":["bzxc3uBcwMv3h47xE","FmMWikxn3hFTmELuA","6AGHfbf4TXye2tqMA"]}