introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ml_2016/lecture/introduction...
TRANSCRIPT
![Page 1: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/1.jpg)
Introduction of this course李宏毅
Hung-yi Lee
![Page 2: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/2.jpg)
Welcome our TAs
![Page 3: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/3.jpg)
林資偉
![Page 4: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/4.jpg)
盧柏儒
![Page 5: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/5.jpg)
方為
![Page 6: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/6.jpg)
宋昀蓁
![Page 7: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/7.jpg)
賴顗安
![Page 8: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/8.jpg)
沈家豪
![Page 9: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/9.jpg)
林賢進
![Page 10: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/10.jpg)
敖家維
![Page 11: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/11.jpg)
吳柏瑜
![Page 12: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/12.jpg)
What are we going to learn?
![Page 13: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/13.jpg)
What is Machine Learning?
“Hi”
“How are you”
“Good bye”
You said “Hello”
A large amount of audio data
You write the program for learning.
Learning ......
![Page 14: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/14.jpg)
“monkey”
“cat”
“dog”
What is Machine Learning?
This is “cat”
A large amount of images
You write the program for learning.
Learning ......
![Page 15: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/15.jpg)
Machine Learning ≈ Looking for a Function• Speech Recognition
• Image Recognition
• Playing Go
• Dialogue System
f
f
f
f
“Cat”
“How are you”
“5-5”
“Hello”“Hi”(what the user said) (system response)
(next move)
![Page 16: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/16.jpg)
Framework
A set of function 21, ff
1f “cat”
1f “dog”
2f “money”
2f “snake”
Model
f “cat”
Image Recognition:
![Page 17: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/17.jpg)
Framework
A set of function 21, ff
f “cat”
Image Recognition:
Model
TrainingData
Goodness of function f
Better!
“monkey” “cat” “dog”
function input:
function output:
![Page 18: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/18.jpg)
Framework
A set of function 21, ff
f “cat”
Image Recognition:
Model
TrainingData
Goodness of function f
“monkey” “cat” “dog”
*f
Pick the “Best” FunctionUsing f
“cat”
Training Testing
Step 1
Step 2 Step 3
![Page 19: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/19.jpg)
Learning Map
Supervised Learning
Regression
Linear Model
Structured Learning
Semi-supervised Learning
Transfer Learning
Unsupervised Learning
Reinforcement Learning
Classification
Deep Learning
SVM, decision tree, K-NN …
Non-linear Model
![Page 20: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/20.jpg)
Learning Map
RegressionThe output of the target
function 𝑓 is “scalar”.
f今天上午 PM2.5
昨天上午PM2.5
…….
明天上午PM2.5
HW1(scalar)
9/01上午 PM2.5 = 63 9/02上午 PM2.5 = 65 9/03上午 PM2.5 = 100
Training Data:
9/12上午 PM2.5 = 30 9/13上午 PM2.5 = 25 9/14上午 PM2.5 = 20
Input:
Input:
Output:
Output:
預測PM2.5
![Page 21: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/21.jpg)
Learning Map
Regression
Classification
![Page 22: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/22.jpg)
Classification
• Binary Classification • Multi-class Classification
Function f Function f
Input Input
Yes or No Class 1, Class 2, … Class N
![Page 23: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/23.jpg)
Binary Classification
Spam filtering
(http://spam-filter-review.toptenreviews.com/)
Function Yes/No
Yes
No
HW2
TrainingData
![Page 24: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/24.jpg)
Multi-class Classification
http://top-breaking-news.com/
Function
政治
體育
經濟
體育 政治 財經
Training Data
DocumentClassification
![Page 25: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/25.jpg)
Learning Map
Regression
Linear Model
Classification
Deep Learning
Non-linear Model
![Page 26: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/26.jpg)
Classification - Deep Learning
• Image Recognition
Function
“monkey”
“cat”
“dog”
“monkey”
“cat”
“dog”
Training Data
Each possible object is a classHW3
Convolutional Neural Network (CNN)
![Page 27: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/27.jpg)
Classification - Deep Learning
• Playing GOFunction
(19 x 19 classes)
Next move
Each position is a class
一堆棋譜
Training Data 進藤光 v.s. 社清春
黑: 5之五 白:天元 黑:五之5
![Page 28: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/28.jpg)
Classification - Deep Learning
• Playing GOFunction
(19 x 19 classes)
Next move
Each position is a class
一堆棋譜
Training Data 進藤光 v.s. 社清春
黑: 5之五 白:天元 黑:五之5
Input:黑: 5之五
Output:天元
Input:黑: 5之五、白:天元
Output:五之5
![Page 29: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/29.jpg)
Learning Map
Supervised Learning
Regression
Linear Model
Semi-supervised Learning
Classification
Deep Learning
SVM, decision tree, K-NN …
Non-linear Model
Training Data:
Input/output pair of target function
Function output = label
![Page 30: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/30.jpg)
Semi-supervised Learning
Labelled data
Unlabeled data
cat dog
(Images of cats and dogs)
For example, recognizing cats and dogs
![Page 31: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/31.jpg)
Learning Map
Supervised Learning
Regression
Linear Model
Structured Learning
Semi-supervised Learning
Transfer Learning
Classification
Deep Learning
SVM, decision tree, K-NN …
Non-linear Model
![Page 32: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/32.jpg)
Transfer Learning
Labelled data
cat dog
Data not related to the task considered (can be either labeled or unlabeled)
elephant Haruhi
For example, recognizing cats and dogs
![Page 33: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/33.jpg)
Learning Map
Supervised Learning
Regression
Linear Model
Structured Learning
Semi-supervised Learning
Transfer Learning
Unsupervised Learning
Classification
Deep Learning
SVM, decision tree, K-NN …
Non-linear Model
HW4
![Page 34: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/34.jpg)
http://top-breaking-news.com/
Unsupervised Learning
• Machine Reading: Machine learns the meaning of words from reading a lot of documents without supervision
![Page 35: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/35.jpg)
Unsupervised Learning
Draw something!
Ref: https://openai.com/blog/generative-models/
![Page 36: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/36.jpg)
Unsupervised Learning
[Chung & Lee, INTERSPEECH 2016]
Machine listens to lots of audio book
How about machine watch TV?
![Page 37: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/37.jpg)
Learning Map
Supervised Learning
Regression
Linear Model
Structured Learning
Semi-supervised Learning
Transfer Learning
Unsupervised Learning
Classification
Deep Learning
SVM, decision tree, K-NN …
Non-linear Model
![Page 38: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/38.jpg)
Structured Learning - Beyond Classification
“機器學習”
“大家好,歡迎大家來修機器學習”
“Machine Learning”
f
Speech Recognition
f
Machine Translation
長門
春日
實玖瑠
人臉辨識
![Page 39: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/39.jpg)
Learning Map
Supervised Learning
Regression
Linear Model
Structured Learning
Semi-supervised Learning
Transfer Learning
Unsupervised Learning
Reinforcement Learning
Classification
Deep Learning
SVM, decision tree, K-NN …
Non-linear Model
![Page 40: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/40.jpg)
Reinforcement Learning
![Page 41: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/41.jpg)
Supervised v.s. Reinforcement
• Supervised
• Reinforcement
Hello
Agent
……
Agent
……. ……. ……
Bad
“Hello” Say “Hi”
“Bye bye” Say “Good bye”Learning from
teacher
Learning from critics
![Page 42: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/42.jpg)
Supervised v.s. Reinforcement
• Supervised:
• Reinforcement Learning
Next move:“5-5”
Next move:“3-3”
First move …… many moves …… Win!
Alpha Go is supervised learning + reinforcement learning.
![Page 43: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/43.jpg)
Learning Map
Supervised Learning
Regression
Linear Model
Structured Learning
Semi-supervised Learning
Transfer Learning
Unsupervised Learning
Reinforcement Learning
Classification
Deep Learning
SVM, decision tree, K-NN …
Non-linear Model
scenario methodtask
![Page 44: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/44.jpg)
Why I need to learn Machine Learning?
![Page 45: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/45.jpg)
AI即將取代多數的工作?
• New Job in AI Age
http://www.express.co.uk/news/science/651202/First-step-towards-The-Terminator-becoming-reality-AI-beats-champ-of-world-s-oldest-game
AI 訓練師(機器學習專家、資料科學家)
![Page 46: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/46.jpg)
AI 訓練師
機器不是自己會學嗎?為什麼需要 AI訓練師
戰鬥是寶可夢在打,為什麼需要寶可夢訓練師?
![Page 47: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/47.jpg)
AI 訓練師
寶可夢訓練師• 寶可夢訓練師要挑選適合的寶可夢來戰鬥
• 寶可夢有不同的屬性
• 召喚出來的寶可夢不一定聽話
• E.g. 小智的噴火龍
• 需要足夠的經驗
AI 訓練師• 在 step 1,AI訓練師要挑選合適的模型
• 不同模型適合處理不同的問題
• 不一定能在 step 3 找出best function
• E.g. Deep Learning
• 需要足夠的經驗
![Page 48: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/48.jpg)
AI 訓練師
•厲害的 AI, AI訓練師功不可沒
•讓我們一起朝 AI訓練師之路邁進
http://www.gvm.com.tw/webonly_content_10787.html
![Page 49: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/49.jpg)
Policy
![Page 50: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/50.jpg)
上課教材
•以後上課會錄音
•上課投影片和錄音會放到 ceiba和李宏毅的個人網頁上
•李宏毅的個人網頁:http://speech.ee.ntu.edu.tw/~tlkagk/courses_ML16.html
![Page 51: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/51.jpg)
FB 社團
•社團: Machine Learning (2016, Fall)• https://www.facebook.com/groups/1774853276101781
/
•有問題可以直接在 FB社團上發問
•如果有同學知道答案請幫忙回答
•有想法也可以在 FB社團上發言
•會紀錄好的問題、答案、留言,期末會加分
![Page 52: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/52.jpg)
評量方式
•不點名、不考試
•作業:沒有分組、每個人都要繳交• 作業一 (10%):9/30 – 10/14 (二週)
• 作業二 (10%):10/14 – 10/28 (二週)
• 作業三 (10%):10/28 – 11/18 (三週)
• 作業四 (10%):11/18 – 12/02 (二週)
•期末專題 (60%):分組進行
•以比賽方式進行
![Page 53: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/53.jpg)
評量方式 -期末專題
• 11/18公告
• 2 ~ 4人一組
•進行方式:會公告幾個可能的題目給同學們選擇
• Intrusion Detection比賽
• Fintech 比賽 (規劃中)
•指定的 Kaggle比賽
•最後會有組內互評
![Page 54: Introduction of this coursespeech.ee.ntu.edu.tw/~tlkagk/courses/ML_2016/Lecture/Introduction (v4).pdf · (19 x 19 classes) Next move Each position is a class 一堆棋譜 Training](https://reader036.vdocuments.site/reader036/viewer/2022062507/5fd15e59959b8238bf0f9bdf/html5/thumbnails/54.jpg)
加簽
•助教會公告作業 0 ,今天晚上 10:00前完成
•作業 0跟機器學習無關,只是測驗基礎程式能力
•如果可以解決空間的問題,完成作業 0 就加簽,助教會公告授權碼取得方式
•但如果無法解決,就不得不有所篩選
•這學期沒修到也不用太難過,未來還會再開