当前位置：首页 > news >正文

【专题】深度学习期末复习资料（题库）

news 来源：原创 2025/6/1 8:23:51

深度学习期末复习资料（题库）

链接：https://blog.csdn.net/Pqf18064375973/article/details/148322500?sharetype=blogdetail&sharerId=148322500&sharerefer=PC&sharesource=Pqf18064375973&sharefrom=mp_from_link

【测试】

After this training phase, we only need the —— to sample new (false) realistic data.
A.generator
B.discriminator
C.Networks
D.Adversarial
Which one of the following is the heart of Google’s Machine Learning system and the most popular of the Deep Learning ibraries ?
A.Caffe2
B.Theano
C.GraphLab
D.TensorFlow
Can we use GPU for faster computations in TensorFlow?()
A.No, not possible
B.Possible only on cloud
C.Possible only with small datasets
D.Yes, possible
One distinct feature of AlexNet was that it had used various kernels of —— sizes.
A.small
B.same
C.large
D.different
Consider the statements below and choose the correct option:

A : The term tensor refers to the representation of data as multi-dimensional array

B : The term fow refers to the average of operations that one performs on tensors
A.A is true
B.B is true
C.Both A and B are true
D.Both A and B are false
Consider the statements below and choose the correct option:

1.GoogleNet which had won the lLSVRc’14 challenge with an error rate of about 6.7%.
1. GoogleNet had increased the depth using a new type of convolution technigue using the inception module
  A.A is true
  B.B is true
  C.Both A and B are true
  D.Both A and B are false
—— is an example of a full platform because it hosts your Deep Learning applications on a cloud ?
A.Ersatz Labs
B.GraphLab
C.H20 AI
D.deep nets
GraphLab offers three different types of built-in storage .Which one is correct ?()
A.chart.columnar and tabular
B.tabular, columnar, and graph,
C.row,column,tabular
D.tabular, columnar, and chart
Consider the statements below and choose the correct option:

A :Caffe is a pure C++ and CUDA library, which can also be operated in command line, Python, and MatLab interfaces

B : Caffe, as part of Facebook Research and Facebook Open Source, builds upon the original Caffe2 project.
A.A is true
B.B is true
C.Both A and B are true
D.Both A and B are false
Which answer explains better the Flattening?()

A.Transform images to vectors to make it easier to predict

B.Delete unnecessary features to make our dataset cleaner

C.Once we have the pooled feature map, this component transforms the information into a vector. it’s the input we need to cet on with Artificial Neural Networks

D.lt is the last step of CNN.
Which one of the following is the dataset of labels that correspond to x_train?
A.x train
B.x_test
C.y_train
D.y_test
How, deep learning models are built on Keras ?()
A.by using sequential models
B.by using feed_dict
C.by creating place holders and computational graphs
D.by creating data frames
During the process of transfer learning, which one of the following question not be answered ?()
A.What to Transfer
B.When to Transfer
C.Why to Transfer
D.How to Transfer
Below statement shows which of the following equation ?out = tf.add(tf.matmul(X, w), b)
A.Logistic Regression Equaltion
B.Deep ANN equation
C.Random Forest Equation
D.Linear Regression equation
How do you feed external data into placeholders?
A.by using impoar data command
B.by using feed_dict
C.by using read data function
D.Not possible
—— is used To learn a generative model, which describes how data is generated in terms of a probabilistic model?
A.Adversarial
B.Generative
C.Networks
D.discriminator
Consider the statements below and choose the correct option:1. The depth of the stateofthe art neural networks has been steadily increasing (from AlexNet with 152 layers to ResNet with 8 layers)2. The developments in neural net architectures were made possible by sianificant advancements in infrastructure.
A.A is true
B.B is true
C.Both A and B are true
D.Both A and B are false
The accuracy —— over time and the loss —— over time.
A.increases,decreases
B.decreases,increases
C.increases.increases
D.decreases,decreases
Consider the statements below and choose the correct option:

A : Keras is high level deep learning APl written in python for neural networks.

B : Keras supports single backend neural network computations and makes implementing neural networks easy
A.A is true
B.B is true
C.Both A and B are true
D.Both A and B are false
The primary purpose of the —— is to find out which image is from the actual training dataset and which is an output from the generator model
A.generative model
B.discriminator model
C.Networks model
D.None of the above.
The discriminator model takes an example from the doman as —— (real or generated) and poredict a —— class label of real or fake (aenerated)
A.input,binary
B.output,binary
C.input,unary
D.output,unary
What does feed dict do?
A.Feeds external data into computational graphs
B.Creates a new place holder
C.Creates a new tensor
D.Creates a new session
Which of the following is a fuly open-source platform, which is a distributed in-memory ML platform with linear scalability?
A.GraphLab
B.Ersatz Labs
C.H20 AI
D.deep nets
Consider the statements below and choose the correct option:

A : Can you point GraphLab at single data sources in order to train data loads ?

B :GraphLab provides great set of data munging features.
A.A is true
B.B is true
C.Both A and B are true
D.Both A and B are false
Below statement shows which of the following equation ?out = tf.sigmoid(tf.add(tf.matmul(X, w), b))
A.Logistic Regression Equaltion
B.Deep ANN equation
C.Random Forest Equation
D.Linear Regression equation
Identify the correct sequence of building simple Artificial Neural network using Keras ?
```
1)Compile Network   2)Define a Network  3)Fit Network
4)Make Predictions  5)Evaluate Network
```
A.2-3-4-5-1
B.3-4-2-1-5
C.2-3-1-5-4
D.2-1-3-5-4
Consider the statements below and choose the correct option:A : During training, the generator progressively becomes better at creating images that look realB: while the discriminator becomes worse at telling them apart.
A.A is true
B.B is true
C.Both A and B are true
D.Both A and B are false
Consider the statements below and choose the correct option:
A : You can set up each type of model as a service that can not be accessed programmatically through an APl.
B : Theano is a Python library designed for performing mathematical operations on single-dimensional arays and to optimize code complation, pimarly for scientifi
research applications.
A.A is true
B.B is true
C.Both A and B are true
D.Both A and B are false
If the discriminator is welltrained and the generator manages to generate reaHooking images that fool the discriminator,then we have —— ？
A.good discrimintative model
B.good generative model
C.bad discrimintative model
D.bad generative model
A generator ( —— ) learns to create images that look real, while a discriminator( —— ) learns to tell real images apart from fakes
A.The artist, The art critic
B.The art critic, The artist
C.Both of the above
D.None of the above
Consider the statements below and choose the correct option:
1. the key motivator for the ResNet architecture was the observation that,empiricaly,removing layers was not improving the results monotonically.
2. t was counterintuitive because a network with n + 1 ayers should be able to leam at least what a network with n layers could lear, plus something more
  A.A is true
  B.B is true
  C.Both A and B are true
  D.Both A and B are false
The lmageNet dataset itself has about 1.2 million images of …classes.?
A.100
B.1000
C.10000
D.None of the above
Consider the statements below and choose the correct option:

A : A min-max 2-player game between the models where the generator model tries to minimize its loss and maximize the discriminator loss

B : The discriminator model maps the input vector (z) to an output which is similar to the data in the training dataset.
A.A is true
B.B is true
C.Both A and B are true
D.Both A and B are false
Consider the statements below and choose the correct option:

A :The generator is trained, it samples random noise and produces an input from that noise.

B :The input then goes through the generator and gets classifed as either"Real" or"Fake" based on the ability of the discriminator to tell one from the other
O A.Ais true
B.B is true
C.Both A and B are true
D.Both A and B are false
What can be used to trigger retraining or refitting to keep models accurate. ()
A.Bias ldentification
B.Alerts
C.Cataloging
D.All of the above
Consider the statements below and choose the correct option:

A :A platform is a set of tools that users can build on top of .

B : Platforms in other contexts exclude i0S/Android and MacOS/Windows for example.
A.A is true
B.B is true
C.Both A and B are true
D.Both A and B are false
An important feature of the platform is the —— which can be run alongside the deep learning models.
A.GraphLab
B.deep nets
C.GraphLab Canvas
D.Graph Analytics toolset
The ResNet achieved groundbreaking results across several competitions - a. —— % error rate on the lmageNet?
A.3.57
B.2.57
C.5.57
D.All of the above
The —— echanism was the key feature of the ResNet which enabled the training of very deep networks
A.skip connection
B.deep network
C.Batch Normalization
D.None of the above.
InUse deep neural networks as the artificial inteligence (Al)algorithms for training purpose?
A.Adversarial
B.Generative
C.Networks
D.Discriminator

【课后题】

A —— platform provides a set of tools that simplify the process of building a deep net for a custom application.

a) Deep Net

b) Ersatz Labs

c) Dato GraphLab

d) None of the above
H2O.ai and —— are two examples of machine learning software platforms that offer Deep Nets.

a) Ersatz Labs

b) Dato GraphLab

c) Deep Net

d) All of the above.
Which one is the automatically makes highly accurate AI models that classify documents, extract text, tables, and images, and group, label, and refine the extracted information?

a) H2O AI

b) Dato GraphLab

c) Theano

d) TensorFlow
Which TensorFlow was initially released by Google in late —— , though version 1.0.0 was released in early —— data structures?

a) 2015,2017

b) 2017,2015

c) 2015,2016

d) None Of the above
—— is a library based on Python that provides different types of functionality for implementing Deep Learning Models.

a) TensorFlow

b) Keras

c) Linear Regression

d) All of the above
Linear Regression Model is used for predicting the unknown value of a variable (——) from the known value of another variable (——) .

a) Dependent Variable, Independent Variable

b) Independent Variable , Dependent Variable

c) Dependent Variable , Dependent Variable

d) Independent Variable , Independent Variable
A —— encapsulates the control and state of the TensorFlow runtime ?

a) placeholder

b) session

c) variables

d) None of the above
Which one is the following are used which allows your graph to take external inputs as parameters.?

a) session

b) placeholders

c) variables

d) All of the above
Which one of the following function measures how far apart the current output of the model is from that of the desired or target output?

a) activation function

b) Cost function

c) loss function

d) tf.train API
Which one of the following is generally take up the most time in any ML pipeline?

a) Data Visualization

b) Data preprocessing

c) Data post processing

d) Data Analysis
Which one of the following is the dataset of labels that correspond to x_train?

a) x_train

b) x_test

c) y_train

d) y_test
To include categorical dataset in our model, our labels should be converted to………………?

a) one-hot encodings

b) one-hot encoded vector

c) downsample

d) upweigh
The generic way to build a model in Keras is to instantiate a Sequential model and keeps adding —— to it.?

a) keras.models

b) mnist.load_data

c) sns.heatmap

d) keras.layers
The ResNet achieved groundbreaking results across several competitions - a ——% error rate on the ImageNet?

a) 3.57

b) 2.57

c) 5.57

d) All of the above
The —— mechanism was the key feature of the ResNet which enabled the training of very deep networks.

a) skip connection

b) deep network

c) Batch Normalization

d) None of the above.
The ImageNet dataset itself has about 1.2 million images of —— classes.?

a) 1000

b) 100

c) 10000

d) None of the above
‘Freeze’ the initial layers, i.e. use the —— weights and biases that the network has learnt from some other task structures?

a) same

b) Different

c) any

d) None Of the above
A generator (“——”) learns to create images that look real, while a discriminator (“——”) learns to tell real images apart from fakes.

a) The artist, The art critic

b) The art critic, The artist

c) Both of the above

d) None of the above
The discriminator is a —— classification model.

a) normal

b) special

c) real

d) All of the above.
The generator G and the discriminator D are jointly trained in a —— minimax game formulation?

a) two-player

b) one-player

c) Both of the above

d) None of the above.
There is a negligible chance that a region generated in the first step will capture the object entirely. Each region will contain a part of the object. In the end, we just want one region, which captures the object entirely. This statement falls into which of the following subtasks of an object detection problem:

a) Region proposal generation

b) Object classification

c) Non-maximum suppression

d) None of the above.