Gradient Descent

Category: science

The optimization algorithm used to minimize the "error" of the model.

Imagine walking down a misty mountain towards the lowest valley (the lowest error rate). Gradient descent is the algorithm that tells the AI: "Take a small step in the direction of the steepest downward slope." It iteratively improves the model until it hits the bottom (the most accurate output).

Common Examples

• The gradient descent algorithm is the engine that drives our neural network to find the optimal balance for every prediction model.
• If your gradient descent reaches the "valley" too quickly, it might get stuck in a "local minimum" instead of the true global optimal result.