ML/DL Type	Supervised Learning	Unsupervised Learning	Semi-Supervised Learning	Reinforcement Learning	Federated Learning	Deep Learning
Data Requirements	Requireslabeled data for all samples	Does not requirelabeled data	Uses a small amount of labeled data and a large amount of unlabeled data	Learns from interactions with an environment, using rewards and penalties	Uses decentralized data across multiple devices or servers	Requires large amounts of data, labeled or unlabeled depending on the task
Main Functions	Regression, Classification	Clustering, Dimensionality Reduction, Association	Improved classification and regression using both labeled and unlabeled data	Decision-making, Policy Learning	Collaborative learning while preserving data privacy	Feature extraction, Pattern recognition, Complex decision making
Algorithms	KNN, Decision Trees, Random Forests, SVM, Linear/Logistic Regression	K-Means, PCA, Apriori, DBSCAN	Self-training, Co-training, Graph-based methods	Q-Learning, SARSA, Policy Gradient Methods	FedAvg, FedProx, FedMA	CNN, RNN, LSTM, Transformers
Applications	Image classification, Spam detection, Price prediction	Customer segmentation, Anomaly detection, Topic modeling	Text classification with limited labeled data, Image recognition	Game AI, Robotics, Autonomous vehicles	Mobile keyboard prediction, Healthcare data analysis	Natural language processing, Computer vision, Speech recognition
Activation Functions	Varies by algorithm	Varies by algorithm	Varies by algorithm	Varies by algorithm	Varies by algorithm	Sigmoid, ReLU, Tanh, Softmax

Three Steps of Machine Learning:

Supervised Learning

Unsupervised Learning

Semi-Supervised Learning

Reinforcement Learning #

Federated Learning

Transfer Learning

Self-Supervised Learning #Lee Hung-yi

機器學習/深度學習	監督式學習	非監督式學習	半監督式學習	增強式學習	聯邦式學習	深度學習
資料需求	需要標籤(label)所有資料	不需要標籤(label)所有資料	告訴機器少部分正確的資料	不需給機器任何的資料，讓機器從一個正向反饋 (Positive Reward)和負向反饋 (Negative Reward)中學習
功能	回歸 Regression
分類 Classification	群集 Clustering
降維 Dimension reduction
關聯 Association
	Q-Learning ;
TD-Learning
演算法	KNN / Decision Tree / Random Forsest / SVM	K-Means
應用
函式						Sigmoid
ReLu

機器學習三步驟：