Loss: how wrong is the model?

After this lesson you can: Explain what a loss function measures and why training minimises it.

First: Features and labels

Watch

• What is a Loss Function? Understanding How AI Models Learn - IBM Technology, 10 min. Focused beginner-friendly concept video from a trusted channel, directly matching loss functions without requiring code first.
• Loss or Cost Function | Deep Learning Tutorial 11 (Tensorflow Tutorial, Keras & Python) - codebasics, 25 min. Directly targets loss versus cost in a practical deep-learning tutorial at an appropriate beginner concept length.
• Machine Learning Tutorial Python - 4: Gradient Descent and Cost Function - codebasics, 28 min. Useful follow-up connecting cost to gradient descent, showing why training minimizes a numeric error signal.

Notes

A loss function is the model’s score for “how wrong was this prediction?” It compares the model’s output with the true label and returns a number. Small loss means the prediction was close to the answer the training data says is correct. Large loss means it was far away, or confidently wrong. Training uses this number as feedback: change the model so that, across many examples, the loss becomes smaller.

Unlock the full lesson notes, the exercises and the graded checkpoint - ₹5,000 once.