Machine Learning
Random Forest
Input → Output
Input: Features (X) + Labels (y)
Output: Predictions + Feature Importance
Core Process
- Build multiple decision trees
- Each tree votes on prediction
- Majority vote = final prediction
- Average feature importance across trees
Key Parameters
n_estimators: 100-500
max_depth: 10-20
min_samples_split: 2-10Neural Networks
Input → Output
Input: Features (X) + Labels (y)
Output: Predictions + Learned Representations
Core Process
- Forward pass through layers
- Compute loss function
- Backpropagate gradients
- Update weights
Key Parameters
layers: [input, hidden, output]
activation: relu/sigmoid/tanh
learning_rate: 0.001-0.1Support Vector Machine
Input → Output
Input: Features (X) + Labels (y)
Output: Predictions + Support Vectors
Core Process
- Find optimal hyperplane
- Maximize margin between classes
- Use support vectors for boundary
- Apply kernel transformation
Key Parameters
kernel: linear/rbf/poly
C: 0.1-100
gamma: 0.001-1.0Optuna
Input → Output
Input: Parameter space + Objective function
Output: Optimized parameters + Best score
Core Process
- Define parameter search space
- Create objective function
- Run optimization trials
- Return best parameters
Key Parameters
n_trials: 100-1000
direction: minimize/maximize
sampler: tpe/grid/random