SAS Institute A00-405 Online Practice Questions and Exam Preparation

SAS Institute A00-405 Online Questions & Answers

• Question 1:

  Regularization in neural networks represents a set of techniques devised to accomplish what?

  A. Reduce overfitting
  B. Downsample a feature map
  C. Minimize the toss function
  D. Calculate a softmax
  C. Minimize the toss function

• Question 2:

  Which option of the loadlmage action can be used to load images in the subdirectory tree of a directory?

  A. recurse
  B. reload
  C. subdirs
  D. repeat
  D. repeat

• Question 3:

  Which statement is TRUE regarding the impact of node placement"?

  A. When a Concept node appears after a Text Parsing node, only kept terms derived by the Text Parsing node can be used in concept creation
  B. When a Concept node is placed after a Text Parsing node: the term table is affected by the concepts defined in the Concept node and the concepts impact every node that follows the Concept node
  C. When the Concept node appears before the Text Parsing node in the pipeline, the spell- checking features of the Concept node produce parent and child terms that show up as synonyms in the term table created by the Text Parsing node
  D. When a Concept node is placed before a Text Parsing node the term table is affected by the concepts defined in the Concept node and the concepts impact every node that follows the Text Parsing node
  C. When the Concept node appears before the Text Parsing node in the pipeline, the spell- checking features of the Concept node produce parent and child terms that show up as synonyms in the term table created by the Text Parsing node

• Question 4:

  Which feature is enabled in the default settings of the Text Parsing Node?

  A. misspelling detection
  B. synonym list
  C. minimum number of documents
  D. start list
  A. misspelling detection

• Question 5:

  Given an equivalent number of neurons why would you use an LSTM-based recurrent neural network over a conventional recurrent neural network?

  A. Implementation advantages - easier lo parallelize
  B. Size - uses (ewer parameters
  C. Efficiency-requires fewer computations
  D. Sequence memory - learns data dependencies spanning more sequence elements
  B. Size - uses (ewer parameters

• Question 6:

  Refer to the exhibit.

  

  The pixel intensity values for the top left corner of a 224x224 grey scale image are provided in Exhibit A This image is provided as input to the convolutional filter (Exhibit B) with equal zero padding of size 2 on all sides

  What is the first value in the feature map resulting from applying this filler (cross-correlation operation)?

  A. 186
  C. 10
  D. 97
  A. 186

• Question 7:

  Which option is incorporated in the dlTune action?

  A. Bayesian Optimization
  B. Hyperband Method
  C. Genetic Algorithm
  D. Simple Random Sampling
  C. Genetic Algorithm

• Question 8:

  Which activation function should be used to obtain a single scalar with value in (-1,1)9

  A. TanH
  B. Logistic
  C. Softmax
  D. ReLU
  A. TanH

• Question 9:

  Given the code specifying the three initial layers of a convolutional network:

  

  What is the size of me output tensor (image) ot pool?

  A. 10*10
  B. 5*5
  C. 8*8
  D. 4*4
  A. 10*10

• Question 10:

  Refer to the exhibit (width'height'stride, d=depth):

  

  What is the cardinality within this block*?

  A. 4
  B. 32
  C. 3
  D. 5
  D. 5