How does Lasso regression modify the loss function to prevent overfitting in predictive models?

Lasso regression keeps the ordinary least‑squares loss (the sum of squared errors) but adds a penalty that is proportional to the sum of the absolute values of the regression coefficients, multiplied by a tuning parameter λ. Because the penalty grows with the size of the coefficients, the optimizer is forced to keep them small, and when λ is large some coefficients are driven exactly to zero, effectively removing those predictors from the model. This shrinkage reduces the model’s flexibility, so it fits the training data less tightly and is less likely to capture random noise, which is the main cause of overfitting. For example, if a simple two‑feature model has coefficients 3. 2 and –1.