Free Online Dumps for Huawei H13-321-ENU Exam

Posted by: Pdfprep Category: Huawei Online Questions Tags: H13-321-ENU Online Questions Post Date: February 1, 2021

Pdfprep2021-02-01T03:01:31+00:00

H13-321-ENU exam is an important Huawei Certification which can test your professional skills. Candidates want to pass the exam successfully to prove their competence. Our technical experts have collected and certified all of H13-321-ENU questions and answers, which are designed to cover the knowledge points of the HCIP-AI-EI Developer V2.0 H13-321-ENU exam. With these H13-321-ENU prep questions, you can pass the H13-321-ENU easily and go further on Huawei career path.

Page 1 of 7

1. The background difference method is to use the current frame in the image sequence and the background reference model for subtraction, which can realize the detection of moving objects in the image.

TRUE

FALSE

2. What are the traditional voice models?

Recurrent neural network

Convolutional neural network

Gaussian mixture model

Hidden Markov Model

3. What is linguistic content?

Voice

Vocabulary

Text

Grammar

4. The bright part of an image is too bright and the dark part is too dark.

Which technique can be used to improve the visual effect of the image? (Single choice)

Reverse

Gamma correction

Grayscale stretching

Grayscale compression

5. What is not part of the DNN structure?

Dropout

Activation function

Weight

Hidden layer

6. Named entity recognition refers to recognizing entities with specific meanings in the text.

Which ones are mainly included? (Multiple Choice)

Name

Place name

Organization name

Time

Date

7. How many brightness changes can a 4-bit image distinguish? (Single choice)

128

256

8. When the human eye observes the surrounding world, the imaging of the object on the retina is inverted.

TRUE

FALSE

9. Which of the following are hidden units.

maxout unit

Rectifier linear unit

Logistic sigmoid and hyperbolic tangent function

Borel unit

10. Which of the following is not a common method of text vectorization?

CBOW

BDOW

Page 2 of 7

11. What is the spatial adjacency of the two pixels with coordinates (21, 13) and coordinates (22.12)?

There is no connection relationship

Diagonal butt

8_ Adjacent

4-adjacency

12. In Chinese, the part-of-speech of a word is often not fixed. Generally speaking, words with the same homonym and different shapes have different grammars in different scenes, which brings great difficulties to the part-of-speech tagging.

TRUE

FALSE

13. The output of the multi-parameter sensor is a continuous signal, and the digital signal can be processed only after the continuous signal is converted into a discrete signal.

TRUE

FALSE

14. The core idea of convolutional neural network?

Local perception

0 padding

Parameter sharing

Fully connected

15. What algorithm engines are supported by ModelArts training assignments? (Multiple Choice)

PyTorch

MXNet

Spark MLlib

TensorFlow

16. What are the tasks in the voice domain that GMM is suitable for processing? (multiple choices)

Speech recognition

Speech synthesis

Voice noise reduction

Voice tracking

17. What is the correct description of probability and likelihood?

If the random variable X follows a certain distribution, the likelihood refers to the possibility of X = x given the parameters.

If the random variable X follows a certain distribution, the probability refers to the probability of X = x given the parameters

If the random variable X follows a certain distribution, the likelihood refers to the truthfulness of a group of parameters under the condition of X = x

If the random variable X follows a certain distribution, the probability refers to the fact that a certain set of parameters has a truth size under the condition of X=x

18. In the vehicle detection scene of traffic law enforcement, the penalties for vehicles that violate regulations are rather missed, instead of false alarms, so as not to produce false fines, which of the system's instructions are high?

Recall rate

confidence

Accuracy

Correct rate

19. Which of the following convolutional neural networks has the most layers?

ResNet

AlexNet

VGG16

LeNet

20. The bimodal method can automatically obtain the optimal segmentation threshold of the image, which is suitable for all gray image binarization tasks.

TRUE

FALSE

Page 3 of 7

21. Which AI scenarios are Huawei Ascend series chips for?

Private cloud

Edge computing

Consumer terminal

IOT industry terminal

Public cloud

22. (Judgment) In general, Huawei's artificial intelligence development strategy is based on continuous investment in basic research and AI talent training, creating a full-stack full-scenario AI solution and an open global ecology.

TRUE

FALSE

23. Using binary digital segmentation, because of noise, the connected domains of all characters are glued together.

What method can be used to segment characters?

Closed operation

Expansion

Gamma correction

Open operation

24. What is not an acoustic feature of speech?

Frequency

Duration

Amplitude

Semantics

25. Regarding the CBOW and Skip-Gram algorithms, is the following statement incorrect?

Both CBOW and Skip-Gram can train vectors, but skip-Gram is faster than CBOW

CBOW calculates the occurrence probability of a word based on n consecutive words in front of a word or n consecutive words before and after a word

Both the CBOW model and the skip-gram model can be based on the Huffman tree

Skip-Gram is to calculate the probability of several words before and after it according to a certain word

26. (Judgment) The image filter operation is also called template operation. The commonly used template operations include template convolution and template sorting.

TRUE

FALSE

27. What are the main functions of GMM in traditional speech recognition tasks?

The probability that the output feature corresponds to the frame

Probability of output features corresponding to speech

Probability of corresponding state of output feature

Probability of corresponding factors of output characteristics

28. The background difference method is to use the current post in the image sequence and the background reference model for subtraction, which can realize the detection of moving objects in the image.

TRUE

FALSE

29. What are the commonly used image coordinate transformations?

Pan

Zoom

Mirror

Rotate

30. What are the main functions of MoXing? (Single choice)

Deep learning tool library developed by Huawei

Deep learning engine developed by Huawei

Package and acceleration of distributed training acceleration framework and API interface of deep learning engines such as TensorFlow

Simple encapsulation of API interfaces for deep learning engines such as TensorFlow

Page 4 of 7

31. In the GMM-HMM model, GMM is mainly used to find the probability of a certain factor, and HMM is mainly used to model each text-speech pair.

TRUE

FALSE

32. What are the processing procedures included in image digitization? (Multiple choices)

Binarization

Quantification

Grayscale transformation

Sampling

33. What are the difficulties of named entity recognition?

The length is uncertain

The composition of named entities is complex

Numerous types of named entities

Complex nesting situation

34. What are the common color spaces in digital images?

RGB

YUV

CMYK

HSV

35. How many brightness changes can a 4-bit graphic distinguish?

256

128

36. The status sequence of HMM can neither be obtained directly nor through observation sequence.

TRUE

FALSE

37. Image target detection algorithm needs to be completed?

Judgment of target category

Calculation of confidence

Calculation of target edge

Calculation of the target position

38. Which of the following steps in the MFC extraction process?

Add window

Framing

Fourier transform

Cepstrum analysis

39. Which of the following options is not a method in the development of natural language processing?

Methods based on deep learning

Statistical-based methods

Method based on recursion

Rule-based approach

40. In image processing tasks, what is the true label of the data set?

Great Trurh

Great Target

Ground Truth

Ground Target

Page 5 of 7

41. Which network structure does the convolutional neural network include?

Output layer

Convolutional layer

Fully connected layer

Pooling layer

42. What is the difficulty of speech recognition?

Dialect recognition

Multiplayer scene recognition

High noise field identification

Recognition of far-field gram wind

43. In the N-gram model, N refers to ignoring the effect of the above words with a distance greater than N. The larger the N, the richer the sequence information contained in the model, and the amount of calculation increases.

TRUE

FALSE

44. A model in which several linear models are mixed together is called a linear mixed model.

TRUE

FALSE

45. There are a lot of small noises in a photo during storage.

Which operation is the best for noise reduction?

Laplace filter

Gaussian filtering

Median filtering

Mean filtering

46. In natural language processing, the most commonly used deep learning algorithm is CNN, because of its memory function?

TRUE

FALSE

47. Which of the following options is the method of automatically calculating the gradient.

Automatic differentiation

Multivariate function differentiation

Numerical differentiation

Symbolic differentiation

48. Which of the following options is not a commonly used algorithm for keyword extraction? (Single choice)

TF-IDF

TextRank

SSA

LDA

49. What is the correct description about the Gaussian distribution curve?

The Gaussian curve is bell-shaped, with both ends high and the middle low

The smaller the standard deviation, the flatter the curve

The larger the standard deviation, the thinner and taller the curve

The Gaussian curve is bell-shaped, with both ends low and the middle high

50. Which of the following is pretreatment?

Single channel change

Increased contrast

Affine changes

Straight main picture

Page 6 of 7

51. Modifying the H component of the HSV color space will change the image? (Single choice)

Hue

Brightness

Saturation

Contrast

52. When taking light, the face is relatively dark. Use gamma correction to enhance the image to achieve a clear light effect.

Which of the Y limit can be selected. (Single choice)

0.5

53. When photographing a document with a mobile phone, the document in the image is distorted due to the shooting angle. You can find () coordinates of the corresponding key points, () transform the image, and correct the document image.

4, Affine

4, perspective

3, perspective

3, Affine

54. Which of the following templates can achieve the sharpness of the image?

010 111 010

010 131 010

0-10 -14-1 0-10

0-10 -16-1 0-10

55. The accuracy of the target detection position can be evaluated by the IOD value. The iout calculation method is to find the ratio of the union and intersection of the true annotations of the detection results.

TRUE

FALSE

56. Markov chain is a discrete event stochastic process with Markov properties in exponential science.

TRUE

FALSE

57. The request/response of HUAWEI CAI API in accordance with the design of RESTful API includes the following.

Response message header

Request URL

Request message header

Response message body

Request message body

58. Image recognition tasks can be divided into three levels, according to the abstraction of the processing content, from low to high?

Image processing, image analysis, image understanding

Image analysis, image understanding, image processing

Image understanding. Image analysis. Image Processing

Image analysis, image processing, image understanding

59. Use a 3*3 convolution kernel to convolve a three-channel color image.

What is the number of parameters to be learned?

60. An APP that can display plant names and breeding methods after taking photos of plants in pots. Its core is the use of image classification technology.

TRUE

FALSE

Page 7 of 7

61. What is the word segmentation method in the jieba framework?

cut

lcut_for_search

lcut

cut_for_search

62. (Judgment) Back propagation is a common method used to train artificial neural networks.

TRUE

FALSE

63. In the Gaussian mixture model, since the distribution of each sample point is unknown, it is impossible to use the maximum function to solve the parameters.

TRUE

FALSE

64. Dropout is to randomly discard a certain proportion of sample data during model training.

TRUE

FALSE

65. Convolve a pair of three-channel color images with two 3x3 convolution kernels.

How many channels does the resulting feature map have? (Single choice)

66. The histogram operation of the image is to count the distribution frequency of the gray value of the whole image. When the illumination of the image is uneven, the operation effect of the histogram will be affected, and it is necessary to use an improved algorithm for image processing.

TRUE

FALSE

67. What is the main task of syntactic analysis?

Participle

Syntactic structure analysis

Semantic analysis

Dependency analysis

68. Jieba framework is a toolkit for English character processing.

TRUE

FALSE

69. Which of the following options does not belong to the three levels of natural language processing technology?

Semantic analysis

Speech analysis

Syntactic analysis

Lexical analysis

70. What is the description of the standard normal distribution?

Expectation is 0

Expectation is 1

The variance is 0

The variance is 1

Free Online Dumps for Huawei H13-321-ENU Exam

Author