Machine Learning
A field of study in the intersection of data science and artificial intelligence. Use statistical models and learning algorithms to train from data.
- supervised learning: train with labeled data to learn how to predict labels.
- unsupervised learning: train with unlabeled data to learn underlying features and information from the data.
- reinforcement learning: learn desired behavior from experience with reward and punishment.
Interpretability vs. Flexibility
In general as models become more flexible they proportionally lose their ability to be interpreted. Normally in real life situations the interpretation of the effect of features can be just as important as model accuracy.
Tasks:
Supervised:
- Prediction
- Classification: binary or multi-class
- Regression: real variables
Unsupervised:
- Clustering
- Anomaly detection
- Dimensionality reduction
Other:
- Data generation
- Feature selection
Topics
Supervised
- linear regression
- logistic regression
- KNN
- decision tree
- ensemble method
- support vector machine
- Bayes classifier