AI Knowledge

Sample Questions

  1. Which of the following expressions is always equivalent to P(B A)P(A)
    • P(A U B)
    • P(A B)
    • P(A)P(B)
    • P(A And B)
  2. A disease that occurs in 1% of the population has a test with a 3% false positive rate and a 6% false negative rate. If the test comes back positive for a random member of the population, what’s the probability they have the disease?
    • 24%
    • 94%
    • 9%
    • 97%
  3. When using a support vector machine on a training set with two classes, what’s the minimum possible number of support vectors?
    • 0
    • 1
    • 2
    • Depends on the dimension of the input data
  4. Which of the following assumptions does the Naive Bayes algorithm make?
    • The input features are linearly independent
    • The input features are conditionally independent
    • The test data is linearly separable
    • The test data is normally distributed
  5. Which of the following expressions defines the variance of a random variable X?
    • E[X-E[X]^2
    • E[X^2-E[X]]
    • E[X-E[X]^2]
    • E[X-E[X]]
  6. Which if the following is an application of singular value decomposition?
    • Inverting a singular matrix
    • Converting a matrix into Jordan form
    • Finding a low-rank approximation of a matrix
    • Solving an overdetermined system of linear equations
  7. Given the task of finding words which share a common prefix with a given word, which data structure would have the optimal expected asymptotic performance?
    • binary tree
    • sorted array
    • hash set
    • trie
  8. Which of the following approaches will help reduce overfitting?
    • Switch to a model with more parameters
    • Use a smaller training set
    • Reduce the learning rate during gradient descent
    • Add a regularization term to the cost function
  9. Which of the following approaches would give the best tradeoff between speed and accuracy when calculating the product of many probabilities?
    • Use the log-sum-exp trick
    • Use the Fourier transform trick
    • Use a single-precision floating point type(float or float32)
    • Use an arbitrary precision decimal type(BigDecimal in Java)
  10. Which of the following random variables cannot reasonably be considered “llD”?
    • Repeatedly drawing,without replacement,from a deck of cards
    • Repeatedly flipping a fair coin
    • Repeatedly choosing a direction to move during an unbiased random walk
    • Repeatedly flipping a biased coin that comes up heads with probability .7
  11. In a random forest classifier, which of the following choices involves randomness?
    • Choosing which trees to discard during inference
    • Choosing which loss function to use
    • Choosing which branch to follow at a given node
    • Choosing which subset of features to use for a given tree
  12. In general, how does the introduction of “regularization” affect a cost function?
    • It doesn’t affect the cost function at all
    • It ensures that the cost function is convex
    • It penalizes large parameter values
    • It adds a term to account for Gaussian noise