Types of Machine Learning: Supervised, Unsupervised & Reinforcement

Key Characteristics:

• Labeled Data: Input-output pairs available
• Objective: Learn mapping from inputs to outputs
• Two Main Tasks: Classification & Regression
• Evaluation: Accuracy, Precision, Recall, RMSE

Common Algorithms:

• Linear/Logistic Regression
• Random Forests
• Support Vector Machines
• Neural Networks

Python Example:

from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score

# Initialize and train
model = RandomForestClassifier()
model.fit(X_train, y_train)

# Predict and evaluate
preds = model.predict(X_test)
print(f"Accuracy: {accuracy_score(y_test, preds):.2f}")
        

Key Characteristics:

• No Labels: Only input data available
• Objective: Discover hidden patterns
• Main Tasks: Clustering & Dimensionality Reduction
• Evaluation: Silhouette Score, Reconstruction Error

Common Algorithms:

Algorithm	Use Case	Complexity
K-Means	Customer Segmentation	O(n)
DBSCAN	Anomaly Detection	O(n log n)
PCA	Feature Reduction	O(n³)

Python Example:

from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_score

# Cluster data into 3 groups
kmeans = KMeans(n_clusters=3)
clusters = kmeans.fit_predict(X)

# Evaluate
print(f"Silhouette Score: {silhouette_score(X, clusters):.2f}")
        

Key Components:

• Agent: The learning algorithm
• Environment: World the agent interacts with
• Rewards: Feedback signal
• Policy: Agent's behavior strategy

Approaches:

Python Example:

import gym
from stable_baselines3 import PPO

# Create environment
env = gym.make('CartPole-v1')

# Initialize and train agent
model = PPO('MlpPolicy', env, verbose=1)
model.learn(total_timesteps=10000)

# Evaluate
obs = env.reset()
for _ in range(1000):
    action, _ = model.predict(obs)
    obs, _, done, _ = env.step(action)
    if done: break
        

Advanced Combinations:

• Semi-Supervised: Mix of labeled + unlabeled data
• Self-Supervised: Generate labels from data
• Imitation Learning: RL with expert demonstrations
• Multi-Task Learning: Shared representations

Real-World Impact:

Hybrid models achieve 15-30% better performance than single-paradigm approaches in complex tasks like autonomous driving