Discussly

Computer Trivia Questions for College Students — page 3

212 trivia questions with answers, college. Multiple-choice trivia with answers, sorted by topic and level. Reveal the answers below, pick one at random, or present them full-screen.

CollegeTriviaPage 3 of 5
  1. 101.

    Adding more basis functions in a linear model, pick the most probably option:

    • ADoesn't affect bias and variance
    • BDecreases model bias
    • CDecreases variance
    • DDecreases estimation bias
    Answer: Decreases model bias
  2. 102.

    Statement 1| The original ResNets and Transformers are feedforward neural networks. Statement 2| The original Transformers use self-attention, but the original ResNet does not.

    • AFalse, False
    • BTrue, True
    • CFalse, True
    • DTrue, False
    Answer: True, True
  3. 103.

    Based on the paper "Click Trajectories: End-to-End Analysis of the Spam Value Chain", which of the following statements are true? "Spammers" here refer to operators of various parts of the "spam value chain."

    • ASpammers run their spam-advertised web sites on compromised user machines that are part of a botnet.
    • BSpammers need to register domain names in order for their spam-based advertisements to be effective.
    • CB and C
    • DThere is a high cost for spammers to switch acquiring banks.
    Answer: B and C
  4. 104.

    If N is the number of instances in the training dataset, nearest neighbors has a classification run time of

    • AO(1)
    • BO(log N )
    • CO( N )
    • DO( N^2 )
    Answer: O( N )
  5. 105.

    The access matrix approach to protection has the difficulty that

    • Adeciding whether a process has access to a resource is undecidable
    • Bit is not capable of expressing complex protection requirements
    • Cthere is no way to express who has rights to change the access matrix itself
    • Dthe matrix, if stored directly, is large and can be clumsy to manage
    Answer: the matrix, if stored directly, is large and can be clumsy to manage
  6. 106.

    Which of the following is a clustering algorithm in machine learning?

    • AGaussian Naïve Bayes
    • BCART
    • CExpectation Maximization
    • DApriori
    Answer: Expectation Maximization
  7. 107.

    Statement 1| Highway networks were introduced after ResNets and eschew max pooling in favor of convolutions. Statement 2| DenseNets usually cost more memory than ResNets.

    • ATrue, True
    • BFalse, False
    • CTrue, False
    • DFalse, True
    Answer: False, True
  8. 108.

    Statement 1| The softmax function is commonly used in mutliclass logistic regression. Statement 2| The temperature of a nonuniform softmax distribution affects its entropy.

    • AFalse, True
    • BFalse, False
    • CTrue, True
    • DTrue, False
    Answer: True, True
  9. 109.

    Statement 1| If there exists a set of k instances that cannot be shattered by H, then VC(H) < k. Statement 2| If two hypothesis classes H1 and H2 satisfy H1 ⊆ H2, then VC(H1) ≤ VC(H2).

    • AFalse, True
    • BTrue, False
    • CTrue, True
    • DFalse, False
    Answer: False, True
  10. 110.

    Which one of the following is the main reason for pruning a Decision Tree?

    • ATo avoid overfitting the training set
    • BTo make the training set error smaller
    • CTo save computing time during testing
    • DTo save space for storing the Decision Tree
    Answer: To avoid overfitting the training set
  11. 111.

    Statement 1| A neural network's convergence depends on the learning rate. Statement 2| Dropout multiplies randomly chosen activation values by zero.

    • ATrue, False
    • BTrue, True
    • CFalse, False
    • DFalse, True
    Answer: True, True
  12. 112.

    What is a web proxy?

    • AAn agent that makes decisions on the client's behalf when interacting with web applications
    • BA piece of software that makes a web application look like a standalone application, making it easier to test
    • CA piece of software that intercepts and possibly modifies requests (and responses) between a web browser and web server
    • DA simulator for the web, for use when off-line
    Answer: A piece of software that intercepts and possibly modifies requests (and responses) between a web browser and web server
  13. 113.

    Which is true about Batch Normalization?

    • ABatch Normalization is equivalent to Layer Normalization for convolutional neural networks.
    • BAfter applying batch normalization, the layer's activations will follow a standard Gaussian distribution.
    • CThe bias parameter of affine layers becomes redundant if a batch normalization layer follows immediately afterward.
    • DThe standard weight initialization must be changed when using Batch Normalization.
    Answer: The bias parameter of affine layers becomes redundant if a batch normalization layer follows immediately afterward.
  14. 114.

    Consider the representation of six-bit numbers by two's complement, one's complement, or by sign and magnitude. In which representation is there overflow from the addition of the integers 011000 and 011000?

    • ATwo's complement only
    • BSign and magnitude and one's complement only
    • CTwo's complement and one's complement only
    • DAll three representations
    Answer: All three representations
  15. 115.

    Which of the following is true of interrupts?

    • AThey are generated when memory cycles are "stolen".
    • BThey are used in place of data channels.
    • CThey cannot be generated by arithmetic operations.
    • DThey can indicate completion of an I/O operation.
    Answer: They can indicate completion of an I/O operation.
  16. 116.

    Which of the following are benefits of penetration testing?

    • ACompositionality of security properties means tested components are secure even if others change
    • BFull evidence of security: a clean test means a secure system
    • CResults are often reproducible
    • DMakes an adversarial neural network converge more quickly
    Answer: Results are often reproducible
  17. 117.

    What is ethical hacking?

    • AHacking systems (e.g., during penetration testing) to expose vulnerabilities so they can be fixed, rather than exploited
    • B"Hacking" ethics so they justify unintended selfish behavior
    • CHacking into systems run by those whose ethics you disagree with
    • DA slang term for rapid software development, e.g., as part of hackathons
    Answer: Hacking systems (e.g., during penetration testing) to expose vulnerabilities so they can be fixed, rather than exploited
  18. 118.

    Statement 1| The ID3 algorithm is guaranteed to find the optimal decision tree. Statement 2| Consider a continuous probability distribution with density f() that is nonzero everywhere. The probability of a value x is equal to f(x).

    • AFalse, False
    • BTrue, False
    • CTrue, True
    • DFalse, True
    Answer: False, False
  19. 119.

    Which of the following is not a block cipher operating mode?

    • ACBF
    • BECB
    • CCBC
    • DCFB
    Answer: CBF
  20. 120.

    Suppose you want to use fuzzing on a program to try to find memory errors; which of the following statements is true?

    • AYou should not use a grammar-based fuzzer, because its adherence to the grammar means it will not find memory errors
    • BCompiling the program with address sanitizer (ASAN) will make errors harder to reproduce
    • CFuzzing doesn't find memory errors, it finds crashes and hangs
    • DCompiling the program with address sanitizer (ASAN) will make the source of a memory error easier to find
    Answer: Compiling the program with address sanitizer (ASAN) will make the source of a memory error easier to find
  21. 121.

    Let T (n) be defined by T(1) = 7 and T(n + 1) = 3n + T(n) for all integers n ≥ 1. Which of the following represents the order of growth of T(n) as a function of n?

    • AΘ(n)
    • BΘ(n^2)
    • CΘ(n^2 log n)
    • DΘ(n log n)
    Answer: Θ(n^2)
  22. 122.

    The IP protocol is primarily concerned with

    • AReliable delivery of large (multi-packet) messages between machines that are not necessarily directly connected
    • BReliable delivery of packets between directly connected machines
    • CDealing with differences among operating system architectures
    • DRouting packets through the network
    Answer: Routing packets through the network
  23. 123.

    Another term for out-of-distribution detection is?

    • Aone-class detection
    • Banomaly detection
    • Ctrain-test mismatch robustness
    • Dbackground detection
    Answer: anomaly detection
  24. 124.

    Which of the following statements is FALSE about memory reclamation based on reference counting?

    • AReference counting is well suited for reclaiming cyclic structures.
    • BReference counting need not keep track of which cells point to other cells.
    • CReference counting incurs additional space overhead for each memory cell.
    • DReference counting is an alternative to mark-and-sweep garbage collection.
    Answer: Reference counting is well suited for reclaiming cyclic structures.
  25. 125.

    What was the first buffer overflow attack?

    • ALove Bug
    • BMorris Worm
    • CSQL Slammer
    • DCode Red
    Answer: Morris Worm
  26. 126.

    An integer c is a common divisor of two integers x and y if and only if c is a divisor of x and c is a divisor of y. Which of the following sets of integers could possibly be the set of all common divisors of two integers?

    • A{-6, -3, -2, -1, 0, 1, 2, 3, 6}
    • B{-6, -3, -2, -1, 1, 2, 3, 6}
    • C{-6, -2, -1, 0, 1, 2, 6}
    • D{-6,-2, -1, 1, 2, 6}
    Answer: {-6, -3, -2, -1, 1, 2, 3, 6}
  27. 127.

    In Message Condentiality, the transmitted message must make sense to only intended

    • ASender
    • BReceiver
    • CTranslator
    • DModulor
    Answer: Receiver
  28. 128.

    For polynomial regression, which one of these structural assumptions is the one that most affects the trade-off between underfitting and overfitting:

    • AWhether we learn the weights by matrix inversion or gradient descent
    • BThe use of a constant-term unit input
    • CThe polynomial degree
    • DThe assumed variance of the Gaussian noise
    Answer: The polynomial degree
  29. 129.

    Statement 1| The F1 score can be especially useful for datasets with class high imbalance. Statement 2| The area under the ROC curve is one of the main metrics used to assess anomaly detectors.

    • AFalse, False
    • BFalse, True
    • CTrue, True
    • DTrue, False
    Answer: True, True
  30. 130.

    Statement 1| The Stanford Sentiment Treebank contained movie reviews, not book reviews. Statement 2| The Penn Treebank has been used for language modeling.

    • AFalse, False
    • BTrue, False
    • CTrue, True
    • DFalse, True
    Answer: True, True
  31. 131.

    Computational complexity of Gradient descent is,

    • Alinear in D
    • Blinear in N
    • Cdependent on the number of iterations
    • Dpolynomial in D
    Answer: polynomial in D
  32. 132.

    Suppose we like to calculate P(H|E, F) and we have no conditional independence information. Which of the following sets of numbers are sufficient for the calculation?

    • AP(H), P(E|H), P(F|H)
    • BP(E, F), P(E|H), P(F|H)
    • CP(E, F), P(H), P(E|H), P(F|H)
    • DP(E, F), P(H), P(E, F|H)
    Answer: P(E, F), P(H), P(E, F|H)
  33. 133.

    When the data must arrive at the receiver exactly as they were sent, its called

    • AMessage Integrity
    • BMessage Condentiality
    • CMessage Splashing
    • DMessage Sending
    Answer: Message Integrity
  34. 134.

    Which of the following statements about Naive Bayes is incorrect?

    • AAttributes can be nominal or numeric
    • BAttributes are equally important.
    • CAttributes are statistically independent of one another given the class value.
    • DAttributes are statistically dependent of one another given the class value.
    Answer: Attributes are statistically dependent of one another given the class value.
  35. 135.

    For a Gaussian Bayes classifier, which one of these structural assumptions is the one that most affects the trade-off between underfitting and overfitting:

    • AWhether we assume full class covariance matrices or diagonal class covariance matrices
    • BWhether we learn the class centers by Maximum Likelihood or Gradient Descent
    • CWhether we allow classes to have different mean vectors or we force them to share the same mean vector
    • DWhether we have equal class priors or priors estimated from the data.
    Answer: Whether we assume full class covariance matrices or diagonal class covariance matrices
  36. 136.

    When doing least-squares regression with regularisation (assuming that the optimisation can be done exactly), increasing the value of the regularisation parameter λ the testing error.

    • Awill never decrease the training error.
    • Bwill never increase the training error.
    • Cwill never decrease the testing error.
    • Dwill never increase
    Answer: will never decrease the training error.
  37. 137.

    Of the following page-replacement policies, which is guaranteed to incur the minimum number of page faults?

    • AReplace the page whose most recent reference was the longest time in the past.
    • BReplace the page whose next reference will be the longest time in the future.
    • CReplace the page whose most recent reference was the shortest time in the past.
    • DReplace the page whose next reference will be the shortest time in the future.
    Answer: Replace the page whose next reference will be the longest time in the future.
  38. 138.

    Statement 1| The ReLU's gradient is zero for $x<0$, and the sigmoid gradient $\sigma(x)(1-\sigma(x))\le \frac{1}{4}$ for all $x$. Statement 2| The sigmoid has a continuous gradient and the ReLU has a discontinuous gradient.

    • ATrue, True
    • BFalse, True
    • CTrue, False
    • DFalse, False
    Answer: True, True
  39. 139.

    Which of the following guidelines is applicable to initialization of the weight vector in a fully connected neural network.

    • AShould set it to zero in order to preserve symmetry across all neurons
    • BShould set it to zero since otherwise it causes a bias
    • CShould not set it to zero since otherwise it will cause overfitting
    • DShould not set it to zero since otherwise (stochastic) gradient descent will explore a very small space
    Answer: Should not set it to zero since otherwise (stochastic) gradient descent will explore a very small space
  40. 140.

    Statement 1| Maximizing the likelihood of logistic regression model yields multiple local optimums. Statement 2| No classifier can do better than a naive Bayes classifier if the distribution of the data is known.

    • ATrue, False
    • BFalse, True
    • CFalse, False
    • DTrue, True
    Answer: False, False
  41. 141.

    Which of the following comes closest to being a perfectly secure encryption scheme?

    • AEnigma, a transposition cipher
    • BThe Caesar Cipher, a substitution cipher
    • CDES (Data Encryption Standard), a symmetric-key algorithm
    • DOne-time pad
    Answer: One-time pad
  42. 142.

    Why apps developed in languages like C, C++ is prone to Buffer-overflow?

    • ANo database check
    • BNo string boundary checks in predefined functions
    • CNo processing power check
    • DNo storage check in the external memory
    Answer: No string boundary checks in predefined functions
  43. 143.

    Statement 1| RELUs are not monotonic, but sigmoids are monotonic. Statement 2| Neural networks trained with gradient descent with high probability converge to the global optimum.

    • AFalse, False
    • BTrue, True
    • CFalse, True
    • DTrue, False
    Answer: False, True
  44. 144.

    How is IP address spoofing detected?

    • AComparing the TTL values of the actual and spoofed addresses
    • BIdentify all TCP sessions that are initiated but does not complete successfully
    • CInstalling and configuring a IDS that can read the IP header
    • DImplementing a firewall to the network
    Answer: Comparing the TTL values of the actual and spoofed addresses
  45. 145.

    Many cryptographic protocols base their security on assumptions about the computational difficulty of integer factorization. Integer factorization serves this purpose because we believe that

    • Atesting primality is computationally intractable
    • Binteger multiplication is a function whose inverse, factorization, remains difficult for a large class of inputs
    • Ceven if P = NP, integer factorization is still likely not to be polynomial-time computable
    • DP = NP
    Answer: integer multiplication is a function whose inverse, factorization, remains difficult for a large class of inputs
  46. 146.

    The K-means algorithm:

    • AConverges to the global optimum if and only if the initial means are chosen as some of the samples themselves
    • BRequires the dimension of the feature space to be no bigger than the number of samples
    • CMinimizes the within class variance for a given number of clusters
    • DHas the smallest value of the objective function when K = 1
    Answer: Minimizes the within class variance for a given number of clusters
  47. 147.

    Local caching of files is common in distributed file systems, but it has the disadvantage that

    • ATemporary inconsistencies among views of a file by different machines can result
    • BA much higher amount of network traffic results
    • CCaching makes file migration impossible
    • DThe file system is likely to be corrupted when a computer crashes
    Answer: Temporary inconsistencies among views of a file by different machines can result
  48. 148.

    Given two Boolean random variables, A and B, where P(A) = 1/2, P(B) = 1/3, and P(A | ¬B) = 1/4, what is P(A | B)?

    • A1/6
    • B3/4
    • C1
    • D1/4
    Answer: 1
  49. 149.

    Statement 1| For any two variables x and y having joint distribution p(x, y), we always have H[x, y] ≥ H[x] + H[y] where H is entropy function. Statement 2| For some directed graphs, moralization decreases the number of edges present in the graph.

    • ATrue, False
    • BFalse, True
    • CTrue, True
    • DFalse, False
    Answer: False, False
  50. 150.

    Suppose you are given an EM algorithm that finds maximum likelihood estimates for a model with latent variables. You are asked to modify the algorithm so that it finds MAP estimates instead. Which step or steps do you need to modify?

    • ABoth
    • BMaximization
    • CExpectation
    • DNo modification necessary
    Answer: Maximization

How to use these trivia questions

  • Warm-up. Open a lesson or meeting with 3–5 questions — enough to wake the room, short enough to keep the agenda.
  • Review game. Pick 10 questions on the topic you just taught and play them in teams; discuss every wrong answer.
  • Quiz night. Mix two or three topics, read each question twice, and give 20 seconds before revealing the answer.

Run them live

Want every person to answer on their own phone, with a timer and a live leaderboard? Use these questions in an EngageSlide Quiz with a live leaderboard — free to start.

Tick questions on any page — your set is kept while you browse (or take the first 10 here) — and they become ready-to-play slides. What is EngageSlide?

More question collections

Questions adapted from MMLU (Hendrycks et al.) (MIT), edited for clarity. This collection may be reused under the same terms — see sources & licenses.