Uncertainty

🐶 용어 정리

• Prediction, Confidence, Probability

🐶Why uncertainty is important?

• Status
  • Before: using the prediction
  • Now: using prediction, uncertainty
• Purpose
  1. Uncertainty inherent in inductive inference
  2. Incorrect model assumptions
  3. noisy or imprecise data
• “... a weather forecaster can be very certain that the chance of rain is 50 %; or her best estimate at 20 % might be very uncertain due to lack of data.” Roughly, the 50 % chance corre- sponds to what one may understand as aleatoric uncertainty, whereas the uncertainty in the 20 % estimate is akin to the notion of epistemic uncertainty. 

🐶 Type of uncertainty 

• Aleatoric (statistical): randomness, variability in the outcome
  • reducible part of the total uncertainty
  • ex. Shannon entropy
• Epistemic (systematic): lack of knowledge (about the best model), uncertainty caused by ignorance 
  • irreducible part of uncertainty

Which type is applicable on my case?

• Sources of uncertainty: Natual trade-off(become solwer to compute as the size of solution space grows)
  1. Noisy sensors through an agent can only partially observe the current state of the world
  2. Dynamics of the environment in response to an interaction
• Challenges
  

  1. Leverage Multi-modal: able to decide in which of these different ways the modality should be leveraged / whether be passive, active, interactive