Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
疎行列と Jaccard 類似度の高速計算
Search
na-o-ys
March 29, 2017
Programming
1
630
疎行列と Jaccard 類似度の高速計算
na-o-ys
March 29, 2017
Tweet
Share
More Decks by na-o-ys
See All by na-o-ys
IoTと監視
naoys
1
790
RubyとJIT
naoys
0
160
将棋盤を画像認識したかった
naoys
0
1.5k
Rust で乗り換え案内
naoys
0
630
有理数集合の濃度
naoys
2
130
YARVの最適化について調べた
naoys
0
140
転職会議サービスのAWS移行記録
naoys
0
72
Anonymous Recursion in C++
naoys
0
430
入門AlphaGo
naoys
5
3.8k
Other Decks in Programming
See All in Programming
Go言語はstack overflowの夢を見るか?
logica0419
0
520
TFLintカスタムプラグインで始める Terraformコード品質管理
bells17
2
360
スキーマ駆動で、Zod OpenAPI Honoによる、API開発するために、Hono Takibiというライブラリを作っている
nakita628
0
310
One Enishi After Another
snoozer05
PRO
0
150
『毎日の移動』を支えるGoバックエンド内製開発
yutautsugi
2
280
React Nativeならぬ"Vue Native"が実現するかも?_新世代マルチプラットフォーム開発フレームワークのLynxとLynxのVue.js対応を追ってみよう_Vue Lynx
yut0naga1_fa
2
510
O Que É e Como Funciona o PHP-FPM?
marcelgsantos
0
190
What Spring Developers Should Know About Jakarta EE
ivargrimstad
0
450
The Past, Present, and Future of Enterprise Java
ivargrimstad
0
440
Go言語の特性を活かした公式MCP SDKの設計
hond0413
1
490
なんでRustの環境構築してないのにRust製のツールが動くの? / Why Do Rust-Based Tools Run Without a Rust Environment?
ssssota
13
44k
なぜGoのジェネリクスはこの形なのか? - Featherweight Goが明かす設計の核心
qualiarts
0
250
Featured
See All Featured
Large-scale JavaScript Application Architecture
addyosmani
514
110k
Balancing Empowerment & Direction
lara
5
700
Building a Modern Day E-commerce SEO Strategy
aleyda
44
7.8k
The Pragmatic Product Professional
lauravandoore
36
7k
Building Applications with DynamoDB
mza
96
6.7k
Making the Leap to Tech Lead
cromwellryan
135
9.6k
10 Git Anti Patterns You Should be Aware of
lemiorhan
PRO
657
61k
A better future with KSS
kneath
239
18k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
132
19k
It's Worth the Effort
3n
187
28k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
31
2.7k
How to Ace a Technical Interview
jacobian
280
24k
Transcript
ૄߦྻ ͋Δ͍ Jaccard ྨࣅΛߴͰܭࢉ͢Δํ๏ @na_o_ys
Agenda 1. ૄߦྻͷσʔλߏ 2. Python ͱܭࢉ 3. Python ͱૄߦྻ 4.
Jaccard ྨࣅ
1. ૄߦྻͷσʔλߏ
ૄߦྻͱ ΄ͱΜͲͷཁૉ͕ 0 Ͱ͋Δߦྻ
1. ૄߦྻͷσʔλߏ (1) • ௨ৗͷߦྻ Array • ૄߦྻΛ Array Ͱѻ͏ͱϝϞϦԋࢉແବ
• 0 ϕΫτϧಉ࢜ͷࢉͱ͔໌Β͔ʹແବ
1. ૄߦྻͷσʔλߏ (2) • Compressed Sparse Row (CSR) • CSR
ಉ࢜ͷՃࢉ, ߦྻੵ͕ߴ • ߦϕΫτϧͷऔΓग़͕͠ߴ • ྻϕΫτϧͷऔΓग़͕͠ • (wikipedia)
2. Python ͱܭࢉ
2. Python ͱܭࢉ • ख़ͨ͠ܭࢉϥΠϒϥϦ • NumPy, SciPy • Scikit-learn
ͱ͜ΖͰɺPython ͍ (DEMO)
Python ͍ • 5000 ഒ ࣮ߦ࣌ؒ 1ZUIPO NT Ұ෦/VN1Z NT
શ෦/VN1Z NT
Python-loop is Evil • ߦྻϧʔϓઈରʹॻ͍͍͚ͯͳ͍ • 1 ඵͰऴΘΔͣͷॲཧʹ 2 ͔͔࣌ؒΔ
• ߦϧʔϓ/ྻϧʔϓॻ͔ͳ͍ํ͕ྑ͍ • 1 ඵͰऴΘΔͣͷॲཧʹ 1 ͔͔Δ
3. Python ͱૄߦྻ
3. Python ͱૄߦྻ • scipy.sparse.csr_matrix
ޮతͳߦྻॲཧ • ߦϕΫτϧͷऔΓग़͠ • Ճࢉࢉ, ߦྻੵ • ෦දݱΛ numpy.ndarray ͱͯ͠อ࣋
• औΓग़ͯ͠ૢ࡞Ͱ͖Δ (NumPy ͷੈք Ͱ)
4. Jaccard ྨࣅ
4. Jaccard ྨࣅ • ϕΫτϧಉ࢜ͷྨࣅ • ڠௐϑΟϧλϦϯάͱ͔Ͱ͏ • ϢʔβAͱϢʔβBͲΕ͘Β͍ࣅ͍ͯΔ͔ Jaccard(a,
b) = a・b / (a・a + b・b - a・b)
ࣄͰඞཁʹͳͬͨ͜ͱ • ૄߦྻͷߦϕΫτϧಉ࢜ͷ Jaccard ྨࣅΛ ܭࢉ͍ͨ͠
DEMO
·ͱΊ
·ͱΊ • Python ͍ • ϥΠϒϥϦΛ͏·͘͏ඞཁ͕͋Δ • ϒϩάΛॻ͍ͨ • http://na-o-ys.github.io/others/
2015-11-07-sparse-vector- similarities.html