Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Speaker Deck
PRO
Sign in
Sign up for free
Collaborative Topic Modeling for Recommending Scientific Articles
Shinichi Takayanagi
May 30, 2016
Research
0
1.1k
Collaborative Topic Modeling for Recommending Scientific Articles
論文"Collaborative Topic Modeling for Recommending Scientific Articles"を読んだ際に使用したスライド
Shinichi Takayanagi
May 30, 2016
Tweet
Share
More Decks by Shinichi Takayanagi
See All by Shinichi Takayanagi
[KDD2021 論文読み会] ControlBurn: Feature Selection by Sparse Forests
stakaya
2
1.4k
[ICML2021 論文読み会] Mandoline: Model Evaluation under Distribution Shift
stakaya
0
1.6k
[情報検索/推薦 各社合同 論文読み祭 #1] KDD ‘20 "Embedding-based Retrieval in Facebook Search"
stakaya
2
400
【2020年新人研修資料】ナウでヤングなPython開発入門
stakaya
29
17k
論文読んだ「Simple and Deterministic Matrix Sketching」
stakaya
1
610
Quick Introduction to Approximate Bayesian Computation (ABC) with R"
stakaya
3
220
The Road to Machine Learning Engineer from Data Scientist
stakaya
5
3.3k
論文読んだ「Winner’s Curse: Bias Estimation for Total Effects of Features in Online Controlled Experiments」
stakaya
1
3.6k
論文読んだ「Class Imbalance, Redux」
stakaya
4
2.6k
Other Decks in Research
See All in Research
パーツ探し
yushiku
PRO
1
490
Offline Time-Independent Multi-Agent Path Planning
kei18
0
120
汎用音響信号表現の発展について@TokyoBISHBash#07
daisukelab_cs
0
230
論文紹介 / GeoDiff: A Geometric Diffusion Model for Molecular Conformation Generation
forest1988
1
130
アノテーションのバイアス排除に関する2020年代の研究動向
kuri8ive
1
490
蓄電池経済効果シミュレーションを提示するとお客さんは販売会社を信頼するか?購入意欲はアップするか?最新調査結果
satoru_higuchi
0
1.9k
Navigation for a team of agents
kei18
0
370
企業の女性活躍推進 分析レポート2022年版 -Forbes JAPAN WOMEN AWARD
libinc
1
180
Fairseq初心者向けチュートリアル
tktkbohshi
2
1.1k
Structure of NLP Papers
ukyh
2
130
P値のトリセツ
taka88
10
5.3k
AI最新論文読み会2022年6月
ailaboocu
0
260
Featured
See All Featured
JazzCon 2018 Closing Keynote - Leadership for the Reluctant Leader
reverentgeek
173
8.6k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
12
940
Unsuck your backbone
ammeep
659
55k
Statistics for Hackers
jakevdp
782
210k
Dealing with People You Can't Stand - Big Design 2015
cassininazir
351
21k
Ruby is Unlike a Banana
tanoku
91
9.3k
Code Reviewing Like a Champion
maltzj
506
37k
Put a Button on it: Removing Barriers to Going Fast.
kastner
56
2.3k
It's Worth the Effort
3n
172
26k
YesSQL, Process and Tooling at Scale
rocio
157
12k
Designing for Performance
lara
597
64k
Debugging Ruby Performance
tmm1
65
10k
Transcript
RCO論文輪読会(2016/05/27) “Collaborative topic modeling for recommending scientific articles”(KDD2011) Chong Wang,
David M. Blei 高柳慎一
(C)Recruit Communications Co., Ltd. ABSTRACT 1
(C)Recruit Communications Co., Ltd. 1. INTRODUCTION 2
(C)Recruit Communications Co., Ltd. 1. INTRODUCTION 3
(C)Recruit Communications Co., Ltd. 1. INTRODUCTION 4
(C)Recruit Communications Co., Ltd. 2. BACKGROUND & 2.1 Recommendation Tasks
5
(C)Recruit Communications Co., Ltd. 2.1 Recommendation Tasks 6
(C)Recruit Communications Co., Ltd. 2.1 Recommendation Tasks 7
(C)Recruit Communications Co., Ltd. 2.2 Recommendation by Matrix Factorization 8
(C)Recruit Communications Co., Ltd. 2.2 Recommendation by Matrix Factorization 9
(C)Recruit Communications Co., Ltd. 2.2 Recommendation by Matrix Factorization 10
(C)Recruit Communications Co., Ltd. 2.2 Recommendation by Matrix Factorization 11
(C)Recruit Communications Co., Ltd. 2.3 Probabilistic Topic Models 12
(C)Recruit Communications Co., Ltd. LDAの生成過程 13
(C)Recruit Communications Co., Ltd. LDAの特徴 14
(C)Recruit Communications Co., Ltd. 3. COLLABORATIVE TOPIC REGRESSION 15
(C)Recruit Communications Co., Ltd. COLLABORATIVE TOPIC REGRESSION 16
(C)Recruit Communications Co., Ltd. CTRの生成過程 17
(C)Recruit Communications Co., Ltd. 3. COLLABORATIVE TOPIC REGRESSION 18
(C)Recruit Communications Co., Ltd. CTRのモデルのRegressionたる所以 19
(C)Recruit Communications Co., Ltd. 学習のさせ方 20
(C)Recruit Communications Co., Ltd. 学習のさせ方 21
(C)Recruit Communications Co., Ltd. 簡単な証明 by iPad手書き 22
(C)Recruit Communications Co., Ltd. 学習のさせ方 23
(C)Recruit Communications Co., Ltd. 予測 24
(C)Recruit Communications Co., Ltd. 4. EMPIRICAL STUDY 25
(C)Recruit Communications Co., Ltd. データの規模感 26
(C)Recruit Communications Co., Ltd. 評価 27
(C)Recruit Communications Co., Ltd. 結果 28
(C)Recruit Communications Co., Ltd. 結果 (ライブラリ内の論文数(Fig 5)・ある論文をLikeした数(Fig 6) 依存性) 29
数が増えると Recallが下がる (あまり有名な論文じゃ ないのを出すため) 数が増えると Recallが上がる (みんな見てる論文 だとCFがうまく動く)
(C)Recruit Communications Co., Ltd. 結果(ある2ユーザの好んだトピックを抽出) 30 トピックの潜 在ベクトルの 重みをランキ ングして抽出
(C)Recruit Communications Co., Ltd. 結果(オフセットの大きかった論文BEST 10) 31 ※内容よりもCFが効くケースに相当
(C)Recruit Communications Co., Ltd. 結果(EMの論文がベイズ統計勢にもよく参照されている例) 32 ※内容よりもCFが効く ケースに相当
(C)Recruit Communications Co., Ltd. 結果(逆にトピックが広がらない例) 33 ※内容が支配的なケー スに相当
(C)Recruit Communications Co., Ltd. 5. CONCLUSIONS AND FUTURE WORK 34