Upgrade to Pro — share decks privately, control downloads, hide ads and more …

ML Session n°4

Avatar for Adrien Couque Adrien Couque
April 05, 2017
Technology
0
27

ML Session n°4

Avatar for Adrien Couque

Adrien Couque

April 05, 2017
Tweet

More Decks by Adrien Couque

See All by Adrien Couque
A roadmap to psychohistory
Avatar for Adrien Couque acq
0
56
ML Session n°8
Avatar for Adrien Couque acq
0
54
ML Session n°7
Avatar for Adrien Couque acq
0
20
ML Session n°6
Avatar for Adrien Couque acq
0
26
ML Session n°5
Avatar for Adrien Couque acq
0
48
ML Session n°3
Avatar for Adrien Couque acq
0
77
ML Session n°2
Avatar for Adrien Couque acq
1
69
ML Session n°1
Avatar for Adrien Couque acq
0
68
Intro to Machine Learning for Android developers
Avatar for Adrien Couque acq
0
190

Other Decks in Technology

See All in Technology
AI導入の理想と現実～コストと浸透〜
Avatar for oprst oprstchn
0
110
低レイヤを知りたいPHPerのためのCコンパイラ作成入門 完全版 / Building a C Compiler for PHPers Who Want to Dive into Low-Level Programming - Expanded
Avatar for HASEGAWA Tomoki tomzoh
4
3.3k
Leveraging Open-Source Tools for Creating 3D Tiles in the Urban Environment
Avatar for Simone Giannecchini simboss
PRO
0
120
Should Our Project Join the CNCF? (Japanese Recap)
Avatar for whywaita whywaita
PRO
0
130
標準技術と独自システムで作る「つらくない」SaaS アカウント管理 / Effortless SaaS Account Management with Standard Technologies & Custom Systems
Avatar for Yuya Takeyama yuyatakeyama
3
1.3k
Witchcraft for Memory
Avatar for pocke pocke
1
490
プロダクトエンジニアリング組織への歩み、その現在地 / Our journey to becoming a product engineering organization
Avatar for hiro-torii hiro_torii
0
130
mrubyと micro-ROSが繋ぐロボットの世界
Avatar for Katsuhiko Kageyama kishima
2
360
GeminiとNotebookLMによる金融実務の業務革新
Avatar for abenben abenben
0
240
CI/CD/IaC 久々に0から環境を作ったらこうなりました
Avatar for Kaz Watanabe kaz29
1
190
マーケットプレイス版Oracle WebCenter Content For OCI
Avatar for oracle4engineer oracle4engineer
PRO
3
910
～宇宙最速～ 2025年AWS Summit レポート
Avatar for sleepingsato satodesu
1
1.9k

Featured

See All Featured
Raft: Consensus for Rubyists
Avatar for Patrick Van Stee vanstee
140
7k
Building a Scalable Design System with Sketch
Avatar for Laura Van Doore lauravandoore
462
33k
JavaScript: Past, Present, and Future - NDC Porto 2020
Avatar for David Neal reverentgeek
48
5.4k
What's in a price? How to price your products and services
Avatar for Michael Herold michaelherold
246
12k
Distributed Sagas: A Protocol for Coordinating Microservices
Avatar for Caitie McCaffrey caitiem20
331
22k
[RailsConf 2023 Opening Keynote] The Magic of Rails
Avatar for Eileen M. Uchitelle eileencodes
29
9.5k
The Pragmatic Product Professional
Avatar for Laura Van Doore lauravandoore
35
6.7k
Connecting the Dots Between Site Speed, User Experience & Your Business [WebExpo 2025]
Avatar for Tammy Everts tammyeverts
5
230
Practical Orchestrator
Avatar for Shlomi Noach shlominoach
188
11k
No one is an island. Learnings from fostering a developers community.
Avatar for Antonio thoeni
21
3.3k
Testing 201, or: Great Expectations
Avatar for Joseph Mastey jmmastey
42
7.5k
Large-scale JavaScript Application Architecture
Avatar for Addy Osmani addyosmani
512
110k

Transcript

  1. ML: regularization and neural networks March 2017

  2. Recap

  3. Recap: linear regression

  4. Recap: polynomial regression

  5. Recap: gradient descent

  6. Regularization

  7. Normalization Goal : have all features be equivalent (in size)

    Rescaling : Standardization :

  8. Regularization : goal

  9. Regularization Before : Now : Forces values to stay low

  10. Neural networks

  11. Demo: Tensorflow Playground

  12. Biological neuron

  13. Artificial neuron: Perceptron

  14. Teaching a neuron

  15. Artifical neuron : activation function

  16. Most common activation functions Identity Sigmoid Tanh ReLU

  17. Teaching a neuron

  18. Demo: simple network, but the hard way (92%)

  19. Demo: simple network, but the hard way (92%)

  20. Artificial neuron -> Artificial neural network

  21. Artificial neuron -> Artificial neural network

  22. Teaching a neural net

  23. Demo: Tensorflow Playground

  24. Demo: layered network, still the hard way (97%)

  25. Questions? March 2017