Skip to content

Latest commit

 

History

History
28 lines (20 loc) · 1.5 KB

P v H.md

File metadata and controls

28 lines (20 loc) · 1.5 KB

Parameters vs Hyperparameters

Parameters:

  • Learned from Data: These are the fundamental elements a model learns during training.
  • They directly represent the relationships and patterns discovered within the data.
  • Coefficients in linear regression (weight of each feature)
  • Weights and biases in neural networks (connections between neurons)
  • Decision tree split points.

Hyperparameters:

  • Set Before Training: These are external knobs that control how the model learns from the data.
  • They are not directly learned but are crucial for guiding the learning process.
  • Learning rate (controls how much the model updates its parameters)
  • Number of hidden layers or neurons in a neural network
  • Number of trees in a random forest. Regularization parameters (control model complexity)

Key Differences:

  1. Learning Process: Parameters are learned, while hyperparameters are pre-defined.
  2. Impact: Parameters directly affect the model's predictions, while hyperparameters influence the learning process itself.
  3. Tuning: Parameters are not usually tuned, while hyperparameter tuning is essential to optimize the model's performance.

Choosing the Right Values:

  • Parameters: The model learns the optimal values for its parameters automatically through the training process.
  • Hyperparameters: Finding the best hyperparameters requires experimentation (grid search, random search, etc.) to evaluate the model's performance with different settings.