AI Dictionary › Fondamenti AI

Gradient Descent

Discesa del gradiente

Gradient descent is the optimization method a neural network uses to find parameter values that minimize error. Imagine standing on a mountain in fog, wanting to reach the valley floor: unable to see far, you feel the ground and step in the steepest downhill direction. The gradient is that direction, computed by backpropagation, and the loss function is the altitude to minimize. Repeating small steps, the model gradually descends toward a low-error configuration. The step size, called the learning rate, sets the pace: too large and you overshoot the minimum, too small and progress crawls.

Definition

Variants like stochastic gradient descent or Adam make the process faster and more stable on huge datasets. Understanding it clarifies why training needs many iterations and why choices like the learning rate shape the final result.

Related terms

• Backpropagation
• Loss Function
• Epoch
• Neural Network
• Machine Learning

More in Fondamenti AI

• LLM (Large Language Model)
• Token
• Context Window
• Temperature
• AI Hallucination
• Fine-tuning
• RAG (Retrieval Augmented Generation)
• Embedding

Put it into practice

• Real scenario: apply this concept with an instant AI verdict.
• Daily gym: short drills to train dictionary concepts.
• Playoff game: challenge other players, elimination-style.

📱 Download the Android app (beta) iOS coming soon

Say what you mean. Get what you need.

Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.