What is the step-by-step method to update weights to minimize error?

What is the step-by-step method to update weights to minimize error?

• A. Gradient Descent
• B. Backpropagation
• C. Activation
• D. Normalization

Answer: (A)

Gradient descent is the step-by-step method used to update weights to minimize error. The process centers on moving each weight in the direction that lowers the network’s error: first, you compute how the error changes with respect to each weight (the gradient), and then you adjust the weight by subtracting a fraction of that gradient, controlled by a learning rate. This update is repeated across all weights and through many passes over the data (epochs) until the error stabilizes at a minimum. Backpropagation is the practical way to efficiently compute those gradients for all weights by applying the chain rule through the network. Activation functions introduce nonlinearity and influence how gradients flow, but they don’t perform the weight updates themselves. Normalization scales inputs to improve training stability, but it isn’t the weight-update rule.

Gradient descent is the step-by-step method used to update weights to minimize error. The process centers on moving each weight in the direction that lowers the network’s error: first, you compute how the error changes with respect to each weight (the gradient), and then you adjust the weight by subtracting a fraction of that gradient, controlled by a learning rate. This update is repeated across all weights and through many passes over the data (epochs) until the error stabilizes at a minimum.

Backpropagation is the practical way to efficiently compute those gradients for all weights by applying the chain rule through the network. Activation functions introduce nonlinearity and influence how gradients flow, but they don’t perform the weight updates themselves. Normalization scales inputs to improve training stability, but it isn’t the weight-update rule.