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
Problems of Neural Networks and its solutions
Search
izuna385
June 21, 2018
Technology
0
130
Problems of Neural Networks and its solutions
Residual Connections とBatch Normalizationがメイン
izuna385
June 21, 2018
Tweet
Share
More Decks by izuna385
See All by izuna385
jel: japanese entity linker
izuna385
0
330
Firebase-React-App
izuna385
0
230
React+FastAPIを用いた簡単なWebアプリ作製
izuna385
0
1.6k
UseCase of Entity Linking
izuna385
0
530
Unofficial slides: From Zero to Hero: Human-In-The-Loop Entity Linking in Low Resource Domains (ACL 2020)
izuna385
1
640
Poly-encoders: Transformer Architectures and Pre-training Strategies for Fast and Accurate Multi-sentence Scoring
izuna385
0
810
Zero-shot Entity Linking with Dense Entity Retrieval (Unofficial slides) and Entity Linking future directions
izuna385
3
1k
Entity representation with relational attention
izuna385
0
75
Zero-Shot Entity Linking by Reading Entity Descriptions
izuna385
0
530
Other Decks in Technology
See All in Technology
EMConf JP 2025 懇親会LT / EMConf JP 2025 social gathering
sugamasao
2
200
アジャイルな開発チームでテスト戦略の話は誰がする? / Who Talks About Test Strategy?
ak1210
1
660
技術スタックだけじゃない、業務ドメイン知識のオンボーディングも同じくらいの量が必要な話
niftycorp
PRO
0
120
株式会社Awarefy(アウェアファイ)会社説明資料 / Awarefy-Company-Deck
awarefy
3
11k
Aurora PostgreSQLがCloudWatch Logsに 出力するログの課金を削減してみる #jawsdays2025
non97
1
230
クラウド関連のインシデントケースを収集して見えてきたもの
lhazy
9
1.8k
Amazon Athenaから利用時のGlueのIcebergテーブルのメンテナンスについて
nayuts
0
110
LINEギフトにおけるバックエンド開発
lycorptech_jp
PRO
0
390
わたしがEMとして入社した「最初の100日」の過ごし方 / EMConfJp2025
daiksy
14
5.3k
IoTシステム開発の複雑さを低減するための統合的アーキテクチャ
kentaro
1
120
開発者のための FinOps/FinOps for Engineers
oracle4engineer
PRO
2
210
プルリクエストレビューを終わらせるためのチーム体制 / The Team for Completing Pull Request Reviews
nekonenene
1
210
Featured
See All Featured
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
21
2.5k
Done Done
chrislema
182
16k
Git: the NoSQL Database
bkeepers
PRO
428
65k
jQuery: Nuts, Bolts and Bling
dougneiner
63
7.7k
Improving Core Web Vitals using Speculation Rules API
sergeychernyshev
10
520
Making the Leap to Tech Lead
cromwellryan
133
9.1k
Typedesign – Prime Four
hannesfritz
41
2.5k
The Language of Interfaces
destraynor
156
24k
The World Runs on Bad Software
bkeepers
PRO
67
11k
Designing on Purpose - Digital PM Summit 2013
jponch
117
7.1k
The Power of CSS Pseudo Elements
geoffreycrofte
75
5.5k
Sharpening the Axe: The Primacy of Toolmaking
bcantrill
40
2k
Transcript
1 / 18 Neural Networks
2 / 18 1. NN !
• Residual Network • Batch Normalization 2. 1. • •
3 / 18 Plain NNs(&) ' pros #%
" (ex. CNN, RNN, ...) cons ! $ $
4 / 18 RNN RNN [1] P. Razvan et
al ,"On the difficulty of training recurrent neural networks." International Conference on Machine Learning. 2013. !"#$ !" %"&$ %"#$ %" %"&$ '() '() '() '*+, '*+, -!"# = /(!!"# ) -! -!$# %! : input !! : hidden state '%&' : '() : input / !" = '*+, 2 !"#$ + '() %"
5 / 18 !" !# !$ %" %# %$ &'(
&'( &'( &)*+ &)*+ ,! = .(!! ) ," ,# RNN 3 1, 12 = 1," 12 + 1,# 12 + 1,$ 12 1,$ 12 = 4 "565$ 1,$ 1!$ 7 1!$ 1!6 7 18!6 12 1!$ 1!" = 1!$ 1!# 7 1!# 1!" = &)*+ 9 :;<= >? !# 7 &)*+ 9 :;<= >? !" @A!B @C : !" ~!6E" fix !6
6 / 18 RNN Vanishing/Exploding Gradient : !"#$ !%&
'( )( … … )* '* ………… ………… +( +* !"#$ (-) !%& (-) '% …… '/ )/ +/
7 / 18 ,$+ /' !"#$ !- !"#$ 2 %
× '()* + ×%,- → # !"#$ !"#$ . 2 % × '()*(+).,-×%,- 1%input or 1)* Loss( RNN ."0& Vanishing/Exploding Gradient
8 / 18 +$ DeepNN( ! +
" )*&!/#% ' (→ ! Loss func ! Loss func → Residual Connection, Batch No malization
9 / 18 0), : Residual Connection – -– F(x)
"/#2 → "/ F(x) + x → (4 '$"/ Identity Mapping +%*1&: 3 . ! 3 Identity – [1] He, Kaiming, et al. "Identity mappings in deep residual networks." European Conference on Computer Vision. Springer, Cham, 2016.
10 / 18 : Residual Connection –– ' Forward
$#& Backward !$"& Deep % & input
11 / 18 Residual Connection –– https://icml.cc/2016/tutorials/icml2016_tutorial_deep_residual_networks_kaiminghe.pdf
12 / 18 ResNet Batch Normalization ResNet Residual Block
• ImplementationBatch Normalization NN ! $# • Batch Normalization" ## http://torch.ch/blog/2016/02/04/resnets.html Plain
13 / 18 ( ) 1 2
( ) n … Batch Normalization –Revisit Gaussian-
14 / 18 Batch Normalization -Input Data distribution
- (Convergence) !! Input NN → input
15 / 18 Batch Normalization -distribution - !"#$% & '
= ) & ' ← ' − , - ~/(,, -2) input
16 / 18 Batch Normalization Data distribution •
=(!, ")fix • Batch Normalization Batch Normalization
17 / 18 Batch Normalization – [2]Ioffe, Sergey,
and Christian Szegedy. "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift." (2015). !, # !%$( → normalize scaling '"&# nomalize
18 / 18 DeepNN+ ! /
& -"#.#)%/'( *$ +!→ , Identity – normalize scaling implement Deep Net