AI 2 (Beyond MLE)

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/17

encourage image

There's no tags or description

Looks like no tags are added yet.

18 Terms

1
New cards

What are the two ways represent what we do not know?

Frequentist viewpoint

Bayesian viewpoint

2
New cards

What is frequentist viewpoint?

knowt flashcard image
3
New cards

What is Bayesian viewpoint?

knowt flashcard image
4
New cards

What does posterior probability mean?

The revised probability after taking into account new data

5
New cards

What does a point estimate leave out?

→ Which other values of θ are available

→ Say θ MLE is 0.7, how strongly the data distinguishes 0.65 from 0.75

→ How prior information should affect the estimate

→ How to average over uncertainty when making an estimate

6
New cards

What are two Bayesian estimators to distinguish?


<p></p>
7
New cards

What does a Bayesian analysis do?


<p></p>
8
New cards

What is bayes rule for a parameter?

knowt flashcard image
9
New cards

What is the formula for MAP?

knowt flashcard image
10
New cards

What distinguishes MAP from MLE?

MLE uses the Likelihood, MAP use Likelihood multiplied by the prior.

11
New cards

How do you set up the MAP optimisation?

knowt flashcard image
12
New cards

Derive the MAP optimisation ( g(θ)) to get θ̂MAP?

knowt flashcard image
13
New cards

Why is the stationary point of MAP a maximum?

→ The log posterior is strictly concave therefore its interior stationary point is the unique maximum

14
New cards

Take the negative Logarithm of MAP.

knowt flashcard image
15
New cards

What is Bayesian Interference?

Instead of returning the posterior mode like MAP, it keeps the Posterior Distribution.

16
New cards

What can the posterior distribution help with?

→Where is its peak?

→What is its mean or median?

→What probability does it assign to an interval?

→What is the predictive probability of a future observation?

17
New cards

What are the two types of uncertainty and what do they mean?

Aleatoric: randomness in outcomes

→ Even if the coins bias were known the next flip could still be heads or tails

Epistemic: Uncertainty about the bias

→ With only 10 flips we do not know the value of θ exactly. The posterior represents this uncertainty

18
New cards