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
2019_G検定対策_数学講座01_ディープラーニングでの基礎計算問題/20190125_JDLA_G_Math_1
Search
ITO Akihiro
January 25, 2019
Technology
0
3
2019_G検定対策_数学講座01_ディープラーニングでの基礎計算問題/20190125_JDLA_G_Math_1
G検定対策社内数学講座
--
ディープラーニングでの基礎計算問題
パーセプトロンとCNNと活性化関数
ITO Akihiro
January 25, 2019
Tweet
Share
More Decks by ITO Akihiro
See All by ITO Akihiro
Software + Hardware = Fun++
akit37
0
2
基本的に "リモートしかない" ワーク/20231128_KBS_LT
akit37
0
7
3つの先端技術が コミュニティ軸で融合した話。/20230615_CMCMeetup
akit37
0
4
Bootleg_越境してみたときのアウェイ感。/20230328_CMCMeetup
akit37
0
8
始まりは2017年のG検定。/20221026_AITable
akit37
0
4
kintone知能化計画/20220902_kintone_and_JPStripes
akit37
0
6
外観検査用画像前処理の_コツをコード解説付きで。/20220810_CDLE_LT
akit37
0
6
サブスク課金に銀行振込を追加してみた。その①/20220713_JPStripes
akit37
0
8
CDLE LT会「お試しプログラミング forとifとfunction()」/20200930_CDLE_LT
akit37
0
5
Other Decks in Technology
See All in Technology
iThome2024 Wailing Wall of Enterprise Security
notsurprised
0
280
AWS アーキテクチャ作図入門/aws-architecture-diagram-101
ma2shita
16
6.7k
技術力の伸ばし方を考える
khirata
0
140
Taking Flight with Tailwind CSS
opdavies
0
4.3k
データ基盤を支える技術
chanyou0311
5
2.9k
動画配信サービスのフロントエンド実装に学ぶ設計原則
yud0uhu
0
130
B2C、B2B プロダクトマネジメントの違い(および思考の罠) / B2C, B2B PM and reduction fallacy
ykmc09
5
2.2k
【リラン】AIの光と闇?失敗しないために知っておきたいAIリスクとその対応 ①政府の動き編
tkhresk
0
130
CockroachDB はどのくらい「しぶとい」のか? / How tough is CockroachDB?
kota2and3kan
13
4.9k
TiDBにおけるテーブル設計と最適化の事例
cygames
0
780
パスワードを保存しますか?
hanacchi
0
250
汎用ポリシー言語Rego + OPAと認可・検証事例の紹介 / Introduction Rego & OPA for authorization and validation
mizutani
1
130
Featured
See All Featured
The Mythical Team-Month
searls
217
42k
The MySQL Ecosystem @ GitHub 2015
samlambert
244
12k
Making Projects Easy
brettharned
109
5.5k
How GitHub (no longer) Works
holman
305
140k
A Modern Web Designer's Workflow
chriscoyier
689
190k
Building Effective Engineering Teams - LeadDev
addyosmani
33
1.9k
What’s in a name? Adding method to the madness
productmarketing
PRO
17
2.7k
Designing for humans not robots
tammielis
247
25k
Templates, Plugins, & Blocks: Oh My! Creating the theme that thinks of everything
marktimemedia
20
1.8k
Understanding Cognitive Biases in Performance Measurement
bluesmoon
12
1.1k
How to name files
jennybc
65
94k
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
34
8.9k
Transcript
ディープラーニングでの 基礎計算 〜パーセプトロンとCNNと活性化関数〜 Jun. 2019 created by ITO Akihiro
パーセプトロンの概念 入力 重み バイアス 出力 入力:1 or 0 出力:1 or 0
画像認識の概念 1 2 3 4 5 6 7 8 9
10 11 12 31 32 12 21 22 23 3 1 2 ・ ・ ・ 入力層 中間層 出力層 1 0 1 2 3 4 5 6 7 8 9 10 11 12 0 1 0 0 重み w1-1 ... w1-36 重み w2-1 ... w2-6 それぞれのシグモイドニューロンは、 ある特徴に対して発火する
CNN (Convolutional Neural Network)の概念 畳み込み層 フィル タ3 特徴マップ 3 プーリング層
出力層 ユニット3 特徴を抽出 特徴をさらに濃縮
フィルタを順に当て 画像を単純化し、 特徴を抽出する。 1 2 3 4 5 6 7
8 9 10 11 12 13 14 15 16 17 18 19 20 21 ... ... 121 元画像 フィルタ1 フィルタ2 斜め方向に反応 縦方向に反応
例: 入力画像 (H,W)=(11,11) フィルタサイズ(Fh,Fw)=(3,3) パディング (P)=0 ストライド (S)=2 ⇒ 出力画像 (Oh,Ow)=(5,5) 1 0 0
1 0 1 3 1 2 1 2 2 3 2 3 2 2 2 3 2 1 2 0 1 1 0 0 0 0 0 0 2 1 2 0 2 2 3 2 2 1 2 2 2 1 0 1 0 1 0 公式
活性化関数 ステップ シグモイド tanh (ハイパボリックタンジェント) ReLU ReLU6 0 1 0
0 0 0 6 0 0 1 0 -1 1 0