Machine Learning Prabability
Uncertainty is Normal
Probability provides the language and tools to handle uncertainty.
Probability of an Event
- Event (A): An outcome to which a probability is assigned.
- Sample Space (S): The set of possible outcomes or events.
- Probability Function (P): The function used to assign a probability to an event.
Probability distribution: Shape or distribution of all events in the sample space
Two Schools of Probability
- Frequentist Probability: objective approach
- Events are observed and counted, and their frequencies provide the basis for directly calculating a probability.
- p-values
- confidence intervals
Probability theory was originally developed to analyze the frequencies of events.
- Bayesian Probability: subjective approach
- Probabilities are assigned to events based on evidence and personal belief and are centered around Bayes' theorem.
- Bayes factors
- credible interval for inference
- Bayes estimator
- maximum estimation for parameter estimation
One big advantage of Bayesian interpretation is that it can because to model our uncertainty about events that do not have long-term frequencies.
Joint, Marginal, and Conditional Probability
For two random variables $(A, B)$, probability of event $X = A$ and $Y = B$ equals
- Joint Probability: Probability of events A and B; $P(A \ and \ B) = P(A\cap B)= P(X=A, Y=B)$
- Marginal Probability: Probability of event A given variable Y; $P(X = A) = \sum_{i \in Y} P(X=A, Y=i)$
- Conditional Probability: Probability of event A given event B; $\displaystyle P(X=A|Y=B) = \frac{P(X=A, Y=B)}{P(Y=B)}$
Probability for Indepencdence and Exclusivity
- Independent $X$ and $Y$
- Joint: $P(X=A \cap Y=B) = P(X=A) P(Y=B)$
- Marginal: $P(X=A, Y) = P(X=A)$
- Conditional: $P(X=A|Y=B) = P(X=A)$
- Mutually exclusive $X$ and $Y$
- Joint: $P(X=A \cap Y=B) = 0
- $P(A \ or \ B) = P(A) + P(B) = P(A \cup B)$