Code Archive Skip to content Google About Google Privacy Terms
去年の終わりから、バイナリハッシングを使った近似近傍検索をいろいろ調べていたのですが、ぼちぼち一段落したので、ひと通りまとめておきます。 バイナリハッシングとは。 個の 次元の点からなるデータセット で、元空間での近傍点を、類似したバイナリコードに関連づける技術。 要するに、実数ベクトルの検索をマトモにやるには、最近のデータは膨大すぎるのでお手上げ。なので、元空間での距離をなるべく保ったまま、バイナリコードに落としましょう。 そうすると、バイナリ一致か、1ビット違うか、2ビット違うか...と、捜索していくにしても、元空間のデータでやるより高速で、しかもストレージ容量を削減できるというわけです。 その ビットのバイナリコード を作るために、 個のハッシュ関数が使われる。 ハッシュ関数は と定義される。ここで、 はデータセット。 は射影ベクトル。 は閾値。 線形写像ベースのハッシングはシンプル
Using Intel.com Search You can easily search the entire Intel.com site in several ways. Brand Name: Core i9 Document Number: 123456 Code Name: Alder Lake Special Operators: “Ice Lake”, Ice AND Lake, Ice OR Lake, Ice* Quick Links You can also try the quick links below to see results for most popular searches. Product Information Support Drivers & Software
JAVAでデータマイング! 『情報工学の難しいそうなアルゴリズムをJAVAで実装して、ひたすらその結果を公開する』ブログになる予定。 PR Calendar <<March>> S M T W T F S 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Theme NaiveBayes ( 2 ) スムージング ( 0 ) はじめに ( 1 ) 計算テクニック ( 0 ) 外れ値除去 ( 0 ) LSH ( 4 ) 協調フィルタリング ( 0 ) ブースティング ( 0 ) Kmeans ( 0 ) 階層的クラスタリング ( 2 ) EMアルゴリズム ( 0 ) BM ( 0 ) SVD ( 0 ) PLSI ( 0 ) LDA ( 0 ) パーセプトロン ( 0 ) A
昨日のSeasar Conference 2009 Autumnで発表させていただいた『Blogopolisの裏側』の資料を公開します。 Blogopolisの裏側View more documents from kaiseh. 資料の28枚目に、重み付きボロノイ図の重心ベースレイアウトの説明用動画がありました。その動画は以下にアップしました。 講演者の皆さん、運営の皆様、本当にお疲れ様でした! 追記 id:mi-changさん p14ででてる「頂点数」、「多角形数」って何を意味してるんだろう?頂点数が多いということはより多くのタグと結びついているってこと? これは、1つ1つのエントリーやブログ、地区(カテゴリ)に対応する土地の幾何データのことです。例えば、5角形の土地の場合は5個の頂点座標が必要になります。土地の頂点数はレイアウト上の理由で決まるもので、タグとは直接関係はありません。
リリース、障害情報などのサービスのお知らせ
最新の人気エントリーの配信
処理を実行中です
j次のブックマーク
k前のブックマーク
lあとで読む
eコメント一覧を開く
oページを開く