Regarding backpropagation, the following statement is wrong?

Question

Options

• ABackpropagation can only be used in feedforward neural networks
• BBackpropagation can be combined with gradient descent algorithm to update network weights
• CBackpropagation passes through the activation function
• DBack propagation refers to the back propagation of errors through the network

• A
  76% (19)
• B
  4% (1)
• C
  8% (2)
• D
  12% (3)

Explanation

Option A is the wrong statement because backpropagation is not limited to feedforward networks - it is widely used in recurrent neural networks (RNNs), convolutional neural networks (CNNs), and other architectures, making this claim factually incorrect. Options B, C, and D are all true: backpropagation is routinely paired with gradient descent to iteratively update weights by minimizing loss (B); it must pass error gradients back through activation functions using the chain rule to compute how each neuron contributed to the error (C); and by definition, backpropagation literally means propagating error signals backward from the output layer through the network (D). A helpful memory tip: think of the word "only" in option A as a red flag - in machine learning, absolute restrictions like "can only be used in X" are almost always false, since most algorithms have been extended well beyond their original domain.

No community discussion yet for this question.