引導式測驗 · 中等 等級

貝葉斯與頻率統計測驗

測試重複採樣信賴區間、貝葉斯先驗和後驗,以及在指定模型下可信區間的意義。

相關引導路徑貝葉斯與頻率統計
問題 1 的 8

在本指南的框架中,頻率派如何定義機率?

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. 在本指南的框架中,頻率派如何定義機率?
    • As a degree of personal belief that updates with evidence
    • As a fixed number assigned by a regulator
    • As always equal to 0.5 unless proven otherwise
    • As the long-run frequency of an event over repeated trials
  2. 根據本指南,95% 頻率派信賴區間實際聲稱什麼?
    • 95% of such intervals, across repeated experiments, would contain the true value
    • There is a 95% chance the true value lies within this specific interval
    • The parameter is 95% likely to be positive
    • 95% of the data falls within that range
  3. 為什麼貝葉斯可信區間的詮釋與頻率派信賴區間不同?
    • Because it is always narrower
    • Because it never contains the true value
    • Because it can be read as a direct probability statement about the parameter given the model and prior
    • 因為它根本不需要數據
  4. 哪個具名組件是貝葉斯分析必須指定,而標準頻率派最大概似分析不需要的?
    • A p-value threshold
    • A prior distribution
    • A sample size
    • A confidence level
  5. 指南中哪項機器學習技術被描述為貝葉斯推理應用於分類的例子?
    • Logistic regression trained with maximum likelihood
    • Standard gradient descent on a neural network
    • K-均值聚類
    • Naive Bayes classifiers
  6. What must a Bayesian A/B analysis still consider when using results as data arrive over time?
    • That Bayesian analysis makes sample representativeness irrelevant
    • That any stopping rule is automatically valid
    • That a prior and likelihood define the posterior, and the stopping and decision plan still needs care
    • That posterior probabilities remove uncertainty
  7. According to the technical section, what common regularization technique is mathematically equivalent to imposing a Bayesian prior?
    • An L2 penalty
    • 輟學
    • 批量歸一化
    • Early stopping
  8. What approximation methods does the guide mention for computing an intractable Bayesian posterior?
    • Gradient descent and batch normalization
    • Markov Chain Monte Carlo sampling and variational inference
    • Cross-validation and grid search
    • Backpropagation and dropout