Training a Simple Model
How Models Learn
Prediction and error
A model predicts a value from an input . The difference between prediction and truth is the error, often measured with a loss function. Training tries to reduce that loss.
From guess to improvement
At first, a model may guess poorly. With repeated updates, it can learn useful patterns and improve. This is similar to practicing with feedback until performance gets better.
Training Vocabulary
Typical training loop
-
1
Feed input data into the model.
-
2
Compute predictions.
-
3
Measure loss against the true labels.
-
4
Update parameters to reduce loss.
-
5
Repeat for many epochs.
What does loss measure?
Loss quantifies prediction error.
Correct answer: How wrong the model is
What is an optimizer used for?
Optimizers help the model learn from mistakes.
Correct answer: It updates model parameters to reduce loss.
What is one epoch?
An epoch means the model has seen the training set once.
Correct answer: One full pass through the training data