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
SGM: Sequence Generation Model for Multi-Label Classification
Search
onizuka laboratory
October 23, 2018
Research
0
67
SGM: Sequence Generation Model for Multi-Label Classification
弊研究室で行なったCOLING2018読み会の発表資料です。
onizuka laboratory
October 23, 2018
Tweet
Share
More Decks by onizuka laboratory
See All by onizuka laboratory
Phrase-Based & Neural Unsupervised Machine Translation
onilab
0
110
Tell-and-Answer: Towards Explainable Visual Question Answering using Attributes and Captions
onilab
0
64
Card-660: A Reliable Evaluation Framework for Rare Word Representation Models
onilab
0
31
A Word-Complexity Lexicon and A Neural Readability Ranking Model for Lexical Simplification
onilab
0
98
Integrating Transformer and Paraphrase Rules for Sentence Simplification
onilab
0
56
An Auto-Encoder Matching Model for Learning Utterance-Level Semantic Dependency in Dialogue Generation
onilab
0
48
Generating More Interesting Responses in Neural Conversation Models with Distributional Constraints
onilab
0
98
Modeling Multi-turn Conversation with Deep Utterance Aggregation
onilab
0
91
Learning Semantic Sentence Embeddings using Pair-wise Discriminator
onilab
0
110
Other Decks in Research
See All in Research
VAR モデルによる OSS プロジェクト同士が生存性に与える 影響の分析
noppoman
0
130
Trezor Safe 3 ファーストインプレッション
toshihr
0
190
Alternative Photographic Processes Reimagined: The Role of Digital Technology in Revitalizing Classic Printing Techniques【SIGGRAPH Asia 2023】
toremolo72
0
430
CSC590 Lecture 01
javiergs
PRO
0
130
NeurIPS-23 参加報告 + DPO 解説
akifumi_wachi
4
1.5k
Experiments on ROP Attack with Various Instruction Set Architectures
yumulab
0
320
Generative AI - practice and theory
gpeyre
1
560
第12回全日本コンピュータビジョン勉強会:画像の自己教師あり学習における大規模データセット
naok615
0
520
プロシェアリング白書2024_PROSHARING_REPORT_2024
circulation
0
620
20240127_熊本から今いちど真面目に都市交通~めざせ「車1割削減、渋滞半減、公共交通2倍」~ 全国路面電車サミット2024宇都宮
trafficbrain
1
660
Alexander Mielke Hellinger--Kantorovich (a.k.a. Wasserstein-Fisher-Rao) Spaces and Gradient Flows
jjzhu
3
180
Refactoring Mining - The key to unlock software evolution
tsantalis
0
250
Featured
See All Featured
Build your cross-platform service in a week with App Engine
jlugia
225
17k
How to Ace a Technical Interview
jacobian
272
22k
GraphQLとの向き合い方2022年版
quramy
32
12k
The World Runs on Bad Software
bkeepers
PRO
61
6.7k
Faster Mobile Websites
deanohume
299
30k
Fantastic passwords and where to find them - at NoRuKo
philnash
37
2.5k
Bootstrapping a Software Product
garrettdimon
PRO
302
110k
jQuery: Nuts, Bolts and Bling
dougneiner
59
7.1k
Infographics Made Easy
chrislema
238
18k
Statistics for Hackers
jakevdp
789
220k
The MySQL Ecosystem @ GitHub 2015
samlambert
243
12k
Designing for humans not robots
tammielis
248
25k
Transcript
SGM: Sequence Generation Model for Multi-Label Classification 2018/10/23
1
1. 2. 3. 4. 5.
2
1. 2. 3. 4. 5.
3
#" n Multi Label Classification(MLC) (,$. >2 !'& -E n
?6: MLC-E>2Single Label Classification)+ Binary Relevancepairwise ranking loss; $ ! 6: n %0: !C<D3B) ! *574=/ n A3 seq2seq; sequence generation! 18%@-E?9 4
1. 2. 3. 4. 5.
5
6
7
Encoder n Bi-LSTM n $! # n
$! # "# !" = LSTM !"() , +" !" = LSTM !",) , +" !" = !" ; !" 8
9
Attention n * &2#,.+ * '% n Attention *(
3 4" /1 n !" , $" , %"&'-$( decoder!40) 10
Attention n ! n Decoder 11
12
Decoder n LSTM n %#" n !"#$% − 1
! n ( !"#$ % − 1 &! global embedding($) 13
Decoder n $(& !% )' n !" , !$
, %$ n &' !! " # &' ( = * −∞ ( . ) 0 12ℎ456.74 14
Global Embedding n #!* % ) (!* n
#!*-+ ".!* '&/ (exposure bias) n Global embedding $, n !" , !$ ∈ ℝ'×' 15
1. 2. 3. 4. 5.
16
"- n l Reuters Corpus Volume I (RCV1-V2) l
'800,000 ( l Arxiv Academic Paper Dataset (AAPD) l 55,840 )$ !* l &,#.+ % 17
n l Hamming-loss l! ", $ " =
& '()*+( ∑ -./ '()*+(0& 1("- ≠ $ "- ) l Micro-F1 n l Binary Relevance(BR) l Classifier Chains(CC) l Label Powerset(LP) l CNN l CNN-RNN 18
19
1. 2. 3. 4. 5.
20
n Global Embedding ! "#$%
& n 21
n sorting Ablation Experiment
n 22
! 23 n "( )
1. 2. 3. 4. 5.
24
9; n Multi-label classification"68&3/0 =5> ( n 1 decoder2<4
sequence generation 4%@7,)# /' n * ! +. $-:? 25