About Machine learning
⚠️ 해당 내용은 Coursera의 Supervised Machine Learning: Regression and Classification 강의 내용을 정리한 내용입니다.
What is machine learning?
Field of study that gives computers the ability to learn without being explicitly programmed.” Arthur Samuel (1959)
Start from checkers games. learns from tons of games.
Machine learning algorithms
- Supervised learning : use most in real-world appliciations. rapid advancements
- Unsuperviesd learning
- Recommender systems
- Reinforcement learning
In this course, we are gonna learn Superviesed learning, Unsuperviese leaning, and Recommender systems.
- Practical advice for applying learning algorithms : even more important then learning algorithms
Supervised learning
X(input) → y(output label) : Learns from being given “right answers”
applications
| Input(X) | Output(y) | Application |
|---|---|---|
| spam?( 0/1) | spam filtering | |
| audio | text transcripts | speech recognition |
| English | Spanish | machine translation |
| ad, user info | cilck? (0/1) | online advertising |
| image, radar info | position of other cars | self-driving car |
| image of phone | defect? (0/1) | visual inspection |
Regression : Housing price prediction
- How much value will it be if the house size is 750 feet^2? ⇒ predicting by X, y
- Regresison : Predict a number from infinitely many possible outputs
Classification : Breast cancer detection
- malignant : danger of cancer, benign : possible of a just tumor
- Classification : predict categories, small number of possibile outputs
Two or more inputs
- Found a boundary(pink line)
Unsupervised learning
- Data only comes with inputs X, but not output labels y
- Algorithm has to find structure in the data
- Find a similar group or cluster ⇒ Clustering Algorithm
- Find unusual data points ⇒ Anomaly detection
- Compress data using fewer numbers ⇒ Dimensionality reduction
Example of Clustring
-
Google news : keyword ‘panda, twin, zoo’
-
DNA microarray
Clustring : Grouping customers
