1. Trang chủ
  2. » Công Nghệ Thông Tin

100 machine learning mcqs with detailed answer exp

21 1 0
Tài liệu đã được kiểm tra trùng lặp

Đang tải... (xem toàn văn)

Tài liệu hạn chế xem trước, để xem đầy đủ mời bạn chọn Tải xuống

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Tiêu đề 100 Machine Learning MCQs with Detailed Answer Explanations
Trường học Sample University
Chuyên ngành Machine Learning
Thể loại E-book
Năm xuất bản 2023
Thành phố Sample City
Định dạng
Số trang 21
Dung lượng 488,78 KB

Các công cụ chuyển đổi và chỉnh sửa cho tài liệu này

Nội dung

Trang 1

Below is a full set of 100 multiple choice questions covering the basics of machine learning,including the detailed answers and explanations for each one This set is ideal for learners,educators, and anyone seeking to test or reinforce their foundational knowledge in the field.

A Clustering

B Predicting outcomes using labeled data

C Sampling data points

D Reducing database size

K-means clustering groups unlabeled data based on similarity.

A Assigning data to discrete categories

B Estimating a continuous value

C Grouping unlabeled data

D Reducing features

Answer: A

Classification assigns each data point to a specific class.

100 Machine Learning MCQs with Detailed

Answer Explanations

Questions 1 100 with Answers and Explanations

1 What is the main objective of supervised learning?

2 Which of the following is an example of unsupervised learning?

3 Which type of problem is classification?

Trang 2

A Model is too simple

B Model is too flexible and memorizes the data

C Model cannot be trained

5 Which method is mainly used for dimensionality reduction?

6 What do we call data that the model has never seen during training?

7 Which of the following is a supervised algorithm?

8 What does the activation function do in a neural network?

Trang 3

Loss measures the error between predicted outcomes and actual results.

A Mean Squared Error

9 Which is NOT a step in data preprocessing?

10 Which term describes the difference between predicted and actual values?

11 Which evaluation metric is best for classification?

12 What is a hyperparameter?

13 What does K in K Nearest Neighbors represent?

Trang 4

Preprocessing is not a machine learning type but a data preparation method.

A Using all features

B Increasing model depth

D Decision tree regression

14 Which learning type involves a reward system?

15 What is a confusion matrix used for?

16 Which is NOT a type of machine learning?

17 Which technique helps avoid overfitting?

18 Which of these is NOT a regression technique?

Trang 5

B Reduces variance by combining many trees

C Needs little data

SVM can use kernel tricks to operate in high-dimensional feature spaces.

A Convolutional Neural Network

B Recurrent Neural Network

D Rectified Output Curve

19 What is the output of a clustering algorithm?

20 What is the main advantage of random forests?

21 Which machine learning algorithm uses a “kernel trick”?

22 What kind of neural network is mostly used for sequential data?

23 What does ROC stand for in model evaluation?

Trang 6

Dimensionality reduction decreases the number of input variables.

A Increase dataset size

B Detect outliers

C Evaluate model generalization

D Optimize model parameters

Answer: C

Splitting data helps assess model performance on unseen data.

A Mean Absolute Error

B F1 Score

C Mean Squared Error

D R2 Score

Answer: B

F1 Score is used for classification, not regression.

A Makes the model faster

B Prevents underfitting

C Provides robust generalization estimation

D Increases data size

24 Which step reduces features to a lower-dimensional form?

25 What is the purpose of the train-test split?

26 Which measurement is NOT commonly used for regression evaluation?

27 What is cross-validationʼs main benefit?

28 Which method is used to handle missing values?

Trang 7

Decision trees can model non-linear relationships in data.

A Ending the project early

B Stopping training before overfitting

C Pausing during training

D Training one layer at a time

BeautifulSoup is used for web scraping, not ML.

A Data cleaning only

B Creating new input variables

C Data evaluation

D Hyperparameter tuning

29 Ensemble learning combines:

30 Which algorithm works best for non-linear boundaries?

31 What is early stopping?

32 Which library is NOT typically used for ML in Python?

33 What is feature engineering?

Trang 8

KNN stores training data and processes queries at prediction time (“lazy learner”).

A A basic neural unit

B One backward pass

C One full pass over the training data

D A day of training

34 Which one is a feature scaling method?

35 What is “bagging” short for?

36 Which algorithm is lazy in its training phase?

37 A perceptron is:

38 What is an epoch in deep learning?

Trang 9

39 Which metric is suitable for imbalanced datasets?

40 What is the main goal of clustering?

41 Which ML task is best for time series forecasting?

42 Which optimizes neural network training?

43 Which of the following is a loss function for classification?

Trang 10

Overfitting reduces model generalization to new data.

A Proportion of true positives among all positive predictions

B Proportion of true positives among all actual positives

C Proportion of true positives among all negatives

D Proportion of false positives among all predictions

Answer: A

Precision focuses on the accuracy of positive predictions.

A Creates new labels

B Converts categorical values to numbers

C Removes null values

44 Which is NOT a clustering algorithm?

45 Which method does NOT improve model generalization?

46 What is precision?

47 What does “label encoding” do?

48 A ROC curve closer to the top left signifies:

Trang 11

49 Which method reduces the number of features but keeps variance?

50 Which step is NOT part of machine learning workflow?

51 Which machine learning library supports deep learning?

52 Which algorithm is best for text classification?

53 Which model parameter is adjusted during training?

Trang 12

Decision trees are prone to overfitting small or noisy datasets.

A Categorical features into multiple binary features

Decision trees use a greedy approach at split points.

A Model memorizes data

B Model canʼt capture data structure

C Model is too flexible

D Too many features

54 Which algorithm is likely to overfit small datasets?

55 A one-hot encoding expands:

56 Which is a greedy algorithm?

57 What does “underfitting” mean?

58 Which type of learning requires expert labeling?

Trang 13

Isolation Forest is designed for detecting anomalies or outliers in data.

A Remove irrelevant features

B Add more features

59 What is a “decision boundary”?

60 Which method is used for anomaly detection?

61 What does feature selection do?

62 Which is NOT an ensemble method?

63 What splits data into k subsets, trains and validates k times?

Trang 14

Regression predicts a continuous numeric value.

A Choosing between too simple or too complex models

B Encoding categorical values

C Imputing missing data

D Measuring classification accuracy

The median is less sensitive to outliers compared to the mean.

A Combines weak learners sequentially

B Uses all data at once

C Trains models independently

D Uses only regression trees

64 Which is an output of regression?

65 What is the bias-variance tradeoff?

66 Which is NOT an activation function?

67 Which one improves robustness to outliers?

68 Whatʼs a key feature of “boosting”?

Trang 15

Feature importance measures the influence of each input variable on predictions.

A Adds more neurons

B Randomly disables neurons during training

C Removes irrelevant features

D Adds noise to input

Answer: B

Dropout randomly turns off neurons to reduce overfitting.

A Number of data points

B Number of features

C Number of data splits/folds

D Number of hyperparameters

69 Which is a benefit of feature scaling?

70 Which technique handles class imbalance?

71 Which evaluates the importance of each feature?

72 Which best describes dropout in neural networks?

73 What characterizes the k in k-fold cross-validation?

Trang 16

Answer: C

'k' is the number of train/validation splits.

A Makes synthetic data

74 What is a “generator” in machine learning?

75 Which regularization adds all absolute values of weights?

76 Which ML algorithm is based on probability?

77 What is data augmentation?

Trang 17

A Trains one model for each class

B Trains one model for all classes

C Trains only on minority class

D Trains two models for each class

Batch size defines how many samples are used for each parameter update.

A Training from scratch

B Reusing pretrained model knowledge

Word embeddings represent text in numerical vector form.

A Batch gradient descent

B Stochastic gradient descent

C Cross-validation

D Overfitting

Answer: B

SGD updates parameters for each sample, increasing speed with big data.

78 What does “one-vs-rest” strategy do?

79 What is batch size in deep learning?

80 What is Transfer Learning?

81 Whatʼs the main use of “word embeddings”?

82 Which increases training speed with large data?

Trang 18

A Size of training set

B Step size in parameter updates

Increasing complexity can cause overfitting; the other options help reduce it.

A Data used for model selection

83 What is the “learning rate”?

84 Which technique is NOT used to reduce overfitting?

85 What is a “blind test set”?

86 Which algorithm is NOT tree-based?

87 What is a “confounding variable”?

Trang 19

Answer: B

A confounder affects both independent and dependent variables, biasing model results.

A More than two classes

B Unequal class distribution

Autoencoders are neural networks for learning compressed representations.

A Sequence of processing steps

B Model weights

C Activation function

D Random variable

88 What does “imbalanced data” mean?

89 What is “feature extraction”?

90 Which Python package is primarily for numerical computing?

91 What is an Autoencoder?

92 What is a “pipeline” in ML?

Trang 20

Silhouette score measures how well data points fit within their clusters.

A Not enough data

B Too many features degrade performance

C Low accuracy

D Underfitting

Answer: B

As feature count increases, data sparsity can negatively impact models.

A Recurrent Neural Network

B Convolutional Neural Network

B Reduces model accuracy

C Reduces computation time

D Adds noise

93 Which is a method for text vectorization?

94 Which metric evaluates clustering quality?

95 What does the “curse of dimensionality” refer to?

96 Which network works best for images?

97 Which is a positive effect of feature selection?

Trang 21

Answer: C

Less features = faster computations and simpler models.

A Final model evaluation

B Tuning model parameters

C Imputing missing values

Larger networks need more data and regularization to avoid overfitting.

You now have a complete 100-question machine learning MCQ set, covering core ideas and providing explanations for each answer.

98 What is the “validation set” for?

99 Which concept refers to applying machine learning to itself?

100 Which is a drawback of large neural networks?

Ngày đăng: 20/07/2025, 21:50

w