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_JD...
Search
Sponsored
·
SiteGround - Reliable hosting with speed, security, and support you can count on.
→
ITO Akihiro
January 25, 2019
Technology
0
9
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
kintone + ローカルLLM = ?
akit37
0
88
【NoMapsTECH 2025】AI Edge Computing Workshop
akit37
0
610
【NoMapsTECH 2025】AI Tech Community Talk
akit37
0
280
エンジニア目線でのテスラ
akit37
0
63
「重鎮問題」について(軽めに)
akit37
0
75
Software + Hardware = Fun++
akit37
0
39
基本的に "リモートしかない" ワーク/20231128_KBS_LT
akit37
1
29
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
30
Bootleg_越境してみたときのアウェイ感。/20230328_CMCMeetup
akit37
0
33
Other Decks in Technology
See All in Technology
All About Sansan – for New Global Engineers
sansan33
PRO
1
1.4k
We Built for Predictability; The Workloads Didn’t Care
stahnma
0
140
Introduction to Sansan for Engineers / エンジニア向け会社紹介
sansan33
PRO
6
68k
Agent Skils
dip_tech
PRO
0
120
Bedrock PolicyでAmazon Bedrock Guardrails利用を強制してみた
yuu551
0
250
仕様書駆動AI開発の実践: Issue→Skill→PRテンプレで 再現性を作る
knishioka
2
680
Why Organizations Fail: ノーベル経済学賞「国家はなぜ衰退するのか」から考えるアジャイル組織論
kawaguti
PRO
1
110
22nd ACRi Webinar - NTT Kawahara-san's slide
nao_sumikawa
0
100
30万人の同時アクセスに耐えたい!新サービスの盤石なリリースを支える負荷試験 / SRE Kaigi 2026
genda
4
1.3k
GitHub Issue Templates + Coding Agentで簡単みんなでIaC/Easy IaC for Everyone with GitHub Issue Templates + Coding Agent
aeonpeople
1
250
学生・新卒・ジュニアから目指すSRE
hiroyaonoe
2
650
Oracle Cloud Observability and Management Platform - OCI 運用監視サービス概要 -
oracle4engineer
PRO
2
14k
Featured
See All Featured
Why Our Code Smells
bkeepers
PRO
340
58k
From Legacy to Launchpad: Building Startup-Ready Communities
dugsong
0
140
Fight the Zombie Pattern Library - RWD Summit 2016
marcelosomers
234
17k
RailsConf 2023
tenderlove
30
1.3k
What does AI have to do with Human Rights?
axbom
PRO
0
2k
What Being in a Rock Band Can Teach Us About Real World SEO
427marketing
0
170
The Spectacular Lies of Maps
axbom
PRO
1
520
What the history of the web can teach us about the future of AI
inesmontani
PRO
1
430
How STYLIGHT went responsive
nonsquared
100
6k
Claude Code どこまでも/ Claude Code Everywhere
nwiizo
61
52k
Product Roadmaps are Hard
iamctodd
PRO
55
12k
The Psychology of Web Performance [Beyond Tellerrand 2023]
tammyeverts
49
3.3k
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