Conditioning (Machine Learning)

Conditioning means providing additional information (like an image or sensor readings) to guide a model’s predictions.

Why It’s Used Instead of learning p(x), we learn p(x∣c), where c is context (e.g., visual input).

  • This lets the model generate outputs relevant to a specific situation.

“Conditioned on”

  • Implies dependence or influence due to past exposure or rules.
  • More common in machine learning, probability, or behavioral science.

Examples:

  • “The model’s output is conditioned on the input image.”
  • “Her behavior was conditioned on past experiences.”

“Conditional on”

  • More formal or academic. Often used in policy, economics, or contract language.
  • Implies that something will happen only if a condition is true.

Examples:

  • “Admission is conditional on passing the entrance exam.”
  • “Funding is conditional on performance metrics.”