引導式測驗 · 中等 等級

貝葉斯網路測驗

測試貝葉斯網絡結構、條件機率、證據更新、分解以及為什麼有向邊單獨不能證明因果關係。

相關引導路徑貝氏網路
問題 1 的 8

Which components define a Bayesian network?

預覽每個問題8 問題

當您參加測驗時,就會出現答案和解釋。

  1. Which components define a Bayesian network?
    • A decision tree with a probability attached to each leaf only
    • A table of observed data with no dependency structure
    • A directed acyclic graph and a conditional probability distribution for each variable given its parents
    • An undirected graph and one global accuracy score
  2. What does a node represent in a Bayesian network?
    • A random variable in the modeled system
    • A database row that always has a known value
    • A proven causal intervention
    • A model-training epoch
  3. Why must a Bayesian-network graph be acyclic?
    • Acyclicity guarantees the graph is causal
    • The model’s dependencies must have an order without a directed cycle
    • Acyclic graphs prevent probabilities from being conditional
    • Cycles make every node independent
  4. How does observing symptoms update a disease probability in a network?
    • It changes graph edges to remove uncertainty
    • It proves the symptom caused the disease
    • It replaces conditional probabilities with a majority vote
    • Inference conditions on observed evidence and computes a posterior distribution
  5. What do local conditional probability tables specify?
    • The distribution of each node given its parent variables
    • The probability that the graph contains a cycle
    • The final model accuracy across all test cases
    • The number of observations stored in a dataset
  6. What does conditional independence allow a Bayesian network to do?
    • Remove the need for probability values
    • Prove the direction of each causal relationship
    • Represent a joint distribution with smaller local factors instead of one full table
    • Guarantee every variable is statistically independent
  7. Does a directed edge in a Bayesian network automatically prove causation?
    • No; directed edges are not allowed in Bayesian networks
    • No; it can encode a factorization or dependence assumption without establishing cause
    • Yes; every directed edge is a confirmed intervention effect
    • Yes, provided the conditional probability table is complete
  8. What should a model report include for a Bayesian-network conclusion?
    • Only the final posterior rounded to a whole percent
    • A statement that the graph is objective because it is visualized
    • The number of edges without how they were specified
    • Variable definitions, graph assumptions, probability sources, and uncertainty in the inference