๐ [Lec1] Neural Networks & Deep Learning
Deep Learning Specialization Course์ ์ฒซ ๋ฒ์งธ ๊ฐ์ 'Neural Networks & Deep Learning'์ 2์ฃผ์ฐจ ๊ณผ์ ์ ๋๋ค.
2์ฃผ์ฐจ ๋ชฉํ
- ์์ ํ, ์ญ์ ํ
- ๋ก์ง์คํฑ ํ๊ท๋ถ์
- ๋น์ฉ ํจ์(Cost Function)
- ๊ฒฝ์ฌํ๊ฐ๋ฒ(gradient descent)
Goal
- Build a logistic regression model structured as a shallow neural network
- Build the general architecture of a learning algorithm, including parameter initialization, cost function and gradient calculation, and optimization implementation (gradient descent)
- Implement computationally efficient and highly vectorized versions of models
- Compute derivatives for logistic regression, using a backpropagation mindset
- Use Numpy functions and Numpy matrix/vector operations
- Work with iPython Notebooks
- Implement vectorization across multiple training examples
Logistic Regression
- Why? - Logistic Regression์ ์ ์ฌ์ฉํ๋ ๊ฑธ๊น?
- What? - Logistic Regression์ด๋ ๋ฌด์์ผ๊น?
- How? - ์ด๋ป๊ฒ ๊ตฌํ ์ ์์๊น?
- Answer
- Why?
- Logistic Regression์ Supervised Learning Problem์ ํด๊ฒฐํ๊ธฐ ์ํด ๊ณ ์๋์๋ค. (Binary Classification)
- What?
- Logistic Regression์ ๋ชฉํ๋ ๋ฌด์์ผ๊น?
- 0๊ณผ 1๋ก ๋ถ๋ฅํ๋ ๋ฌธ์ ๊ฐ ์ฃผ์ด์ก์ ๋(y = {0, 1}) ํด๋์ค๋ฅผ ์์ธกํ๋ ๊ฒ์ด๋ค. e.g ๊ณ ์์ด์ธ๊ฐ vs ๊ณ ์์ด๊ฐ ์๋๊ฐ. ์คํธ์ธ๊ฐ vs ํ์ธ๊ฐ. ๊ฐ์ผ์ธ๊ฐ vs ๊ฐ์ผ์ด ์๋๊ฐ ๋ฑ ๋ค์ํ ์์๋ฅผ ๋ค ์ ์๊ฒ ๋ค
- ํ๋ จ ๋ฐ์ดํฐ์ ๋ชจ๋ธ์ ์์ธก๊ฐ์ ์๋ฌ๋ฅผ ์ค์ด๋ ๊ฒ์ด๋ค.
- How?
- linear function์ sigmoid ํจ์๋ฅผ ์์ ๊ฒฐ๊ณผ๊ฐ์ด 0์์ 1์ฌ์ด๊ฐ ๋๋๋ก ๋ง๋ค์ด์ค๋ค
- Why?
Logistic Regression Cost Function
What is the difference between the cost function and the loss function for logistic regression?
โ Loss Function๊ณผ Cost Fuction์ ์ฐจ์ด๋ฅผ ๋งํ ์ ์์ด์ผํ๋ค
์์ ์์๊ณผ ๊ฐ์ด Loss Function๊ณผ Cost Function์ ๊ฐ์ฅ ํฐ ์ฐจ์ด๋ ์ค์ ๊ฐ๊ณผ ์์ธก๊ฐ์ ์ฐจ์ด๋ฅผ ํ๋์ ๋ฐ์ดํฐ์ ๋ํด์๋ง ์๊ฐํ๋๋ ๋๋ ์ ์ฒด ํธ๋ ์ด๋ ๋ฐ์ดํฐ์ ๋ํด์ ์๊ฐํ๋๋์ ๋ฌ๋ ค์๋ค. ์ ๋ฆฌํ๋ฉด
- Loss Function์ด Single Training Example์ ๋ํ ๊ณ์ฐ์ด๋ผ๋ฉด
- Cost Function์ ์ ์ฒด ํธ๋ ์ด๋ ์ธํธ์ ๋ํ ํ๊ท ๊ฐ์ ๊ณ์ฐํ ๊ฒ์ด๋ค
So the terminology I'm going to use is that the loss function is applied to just a single training example like so. And the cost function is the cost of your parameters. So in training your logistic regression model, we're going to try to find parameters W and B that minimize the overall costs function J written at the bottom.
Derivatives(๋ฏธ๋ถ)
โ ๋ฏธ๋ถ
- ๋ฏธ๋ถ์ ์ง๊ด ⇒ Slope(๊ธฐ์ธ๊ธฐ) ๋ผ๊ณ ์๊ฐํ์
- ํจ์๊ฐ์ด ์ง์ ์ผ ๊ฒฝ์ฐ์๋ ์ด๋ ์์น์์๋ ๋ฏธ๋ถ๊ฐ์ด ๊ฐ์ง๋ง ๊ณก์ ๋๋ ๋ค๋ฅธ ๊ฒฝ์ฐ์๋ ์์น์ ๋ฐ๋ผ ๋ฏธ๋ถ๊ฐ์ด ๋ฌ๋ผ์ง ์ ์๋ค
Logistic Regression Gradient Descent
Logistic Regression on m examples
More Vectorization Examples
Vectorizing Logistic Regression
Reference
โ End! -20.11.03.Tue- :)
โ Update! -20.11.15.Sun am 7:00 - :)
'Coursera > Deep Learning Specialization' ์นดํ ๊ณ ๋ฆฌ์ ๋ค๋ฅธ ๊ธ
[Deep-Special] [Lec1] Week3. Shallow Neural Network (0) | 2020.11.08 |
---|---|
[Deep-Special] [Lec1] Programming Assignment - Logistic Regression with a Neural Network mindset (0) | 2020.11.04 |
[Deep-Special] [Lec1] Week2. Python Basics with numpy(Optional) (0) | 2020.11.04 |
[Deep-Special] [Lec1] Week1. Neural Networks and Deep Learning (0) | 2020.11.02 |
[Coursera] ์์ํ๊ธฐ Financial Aid (5) | 2020.11.02 |
๋๊ธ