While measuring the test coverage of a neural network, a test engineer wants to measure the number of neurons that have each output two activation function results with a minimum difference between…

The correct answer is B. Value-change coverage. Value-change coverage is a coverage measure that focuses on tracking the changes in the values of a neuron's output. In this case, the test engineer is interested in measuring the number of neurons that exhibit a difference of 0.5 between two activation function results, which…

Testing of AI-based Systems

Question

While measuring the test coverage of a neural network, a test engineer wants to measure the number of neurons that have each output two activation function results with a minimum difference between the two results of 0.5. Which ONE of the below coverage measures would achieve that goal?

Options

ANeuron coverage
BValue-change coverage
CSign-change coverage
DNone of the coverage measures

How the community answered

(55 responses)

A
2% (1)
B
84% (46)
C
11% (6)
D
4% (2)

Explanation

Value-change coverage is a coverage measure that focuses on tracking the changes in the values of a neuron's output. In this case, the test engineer is interested in measuring the number of neurons that exhibit a difference of 0.5 between two activation function results, which is effectively a value change. This coverage measure ensures that the neuron undergoes a significant change in its output, fulfilling the engineer's goal.

Topics

#value-change coverage#neural network coverage#activation function#coverage metrics

Community Discussion

No community discussion yet for this question.

Full CT-AI Practice