{"id":4820,"date":"2024-05-07T04:34:00","date_gmt":"2024-05-06T19:34:00","guid":{"rendered":"https:\/\/devneko.jp\/wordpress\/?p=4820"},"modified":"2024-05-07T04:34:00","modified_gmt":"2024-05-06T19:34:00","slug":"kan-kolmogorov-arnold-networks","status":"publish","type":"post","link":"https:\/\/devneko.jp\/wordpress\/?p=4820","title":{"rendered":"KAN: Kolmogorov-Arnold Networks"},"content":{"rendered":"\n<ul class=\"wp-block-list\">\n<li><strong>KAN: Kolmogorov-Arnold Networks&nbsp;<\/strong>[16.8]<br>MLP(Multi-Layer Perceptrons)\u306e\u4ee3\u66ff\u3068\u3057\u3066\u3001KAN(Kolmogorov-Arnold Networks)\u3092\u63d0\u6848\u3059\u308b\u3002 \u30ab\u30f3\u306f\u30a8\u30c3\u30b8\u4e0a\u3067\u5b66\u7fd2\u53ef\u80fd\u306a\u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3\u6a5f\u80fd\u3092\u6301\u3064(&#8220;weights&#8221;)\u3002 \u3053\u306e\u4e00\u898b\u5358\u7d14\u306a\u5909\u5316\u306b\u3088\u308a\u3001KANSA\u306f\u7cbe\u5ea6\u3068\u89e3\u91c8\u53ef\u80fd\u6027\u3068\u3044\u3046\u70b9\u3067\u3001\u30cb\u30e5\u30fc\u30e9\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u3092\u4e0a\u56de\u308a\u307e\u3059\u3002<br><a href=\"http:\/\/arxiv.org\/abs\/2404.19756v1\">\u8ad6\u6587<\/a>&nbsp;&nbsp;<a href=\"https:\/\/fugumt.com\/fugumt\/paper_check\/2404.19756v1\">\u53c2\u8003\u8a33\uff08\u30e1\u30bf\u30c7\u30fc\u30bf\uff09<\/a>&nbsp; &nbsp;(Tue, 30 Apr 2024 17:58:29 GMT)<\/li>\n\n\n\n<li>MLP\u3088\u308a\u3082\u6027\u80fd\u30fb\u89e3\u91c8\u53ef\u80fd\u6027\u304c\u512a\u308c\u3066\u3044\u308b\u3068\u4e3b\u5f35\u3059\u308b\u69cb\u9020\u306e\u63d0\u6848\u3002\u300cKANs and MLPs are dual: KANs have activation functions on edges, while MLPs have activation functions on nodes. This simple change makes KANs better (sometimes much better!) than MLPs in terms of both model accuracy and interpretability.\u300d\u3068\u306e\u3053\u3068\u3002\u73fe\u6642\u70b9\u3067\u306f\u300cCurrently, the biggest bottleneck of KANs lies in its slow training. KANs are usually 10x slower than MLPs, given the same number of parameters.\u300d\u3068\u3044\u3046\u8a18\u8f09\u3082\u3042\u308b\u304c\u3001\u672c\u5f53\u304b\u3064\u5e83\u304f\u53d7\u3051\u5165\u308c\u3089\u308c\u308b\u306e\u3060\u308d\u3046\u304b\u3002\u3002<\/li>\n\n\n\n<li>\u30ea\u30dd\u30b8\u30c8\u30ea\u306f<a href=\"https:\/\/github.com\/KindXiaoming\/pykan\">GitHub &#8211; KindXiaoming\/pykan: Kolmogorov Arnold Networks<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[106,208],"class_list":["post-4820","post","type-post","status-publish","format-standard","hentry","category-arxiv","tag-deep-learning","tag-kan"],"_links":{"self":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4820","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=4820"}],"version-history":[{"count":0,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=\/wp\/v2\/posts\/4820\/revisions"}],"wp:attachment":[{"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=4820"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=4820"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/devneko.jp\/wordpress\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=4820"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}