2.1 An introduction to Bayesian statistics

2.1 Overview BS MED INTRO V2-en

정의

• A statistical inference framework
• Can be used for estimation, classification, detection, model selection, etc.

특징

• unknown quantities are described as random.

활용 예 #1 : 병원

• Quantity of interest: the disease
• Observations: blood samples, temperature, comments by patient, etc
• 결과 : based on our observations, patient has disease X with 97% probability

활용 예 #2 : 자율 주행 차

• Quantity of interest: relative position and velocity of other vehicles at the current time.
• Observations: wheel speeds, INS measurements, radar detections (distance and angle), Lidar point clouds, camera images, etc.
• 결과 : vehicle motions are modelled statistically

기존 통계학과의 차이점 COMPARISON: BAYES VS FREQUENTIST

• We wish to estimate the height of the Eiffel tower.
  • Frequentist perspective: the tower has a certain height and is therefore not random.
  • Bayesian perspective: we describe our uncertainties in the height stochastically -> height is described as random!

OVERVIEW OF THE BAYESIAN STRATEG

• Modeling.
• Measurement update (본 강의에서 주로 다룸)
• Decision making