Why does fitting too much detail to data ruin predictions?
Overfitting
Why does fitting too much detail to data ruin predictions?
Imagine a chef who tries to cook a new recipe by memorizing every step from a single dish instead of learning the basic cooking principles.
The chef's recipe memorization is like overfitting; it works for one dish but fails with others. Overfitting in machine learning is when a model learns the noise in the training data instead of the actual pattern, leading to poor performance on new data.
Example
If the chef's dish was made with 50 unique steps for one meal, it might fail when trying to cook a similar dish with only 10 steps.
Remember this
Overfitting is like memorizing too much, which makes it hard to adapt to new situations.
Text adapted from Wikipedia, licensed under CC BY-SA 4.0.
Related concepts
Deep learning
Why do smart phones sometimes struggle to recognize new faces?
Convolutional neural network
Can a neural network learn too well?
Linear regression
Can we predict something with a straight line?
Ridge regression
Why do straight lines sometimes fail to fit our data perfectly?
batch size affects generalization: larger batches find sharper minima
Larger batch sizes lead to sharper minima, enhancing generalization by providing more accurate gradient estimates
Bias vs variance: high bias = underfitting, high variance = overfitting
Can a perfect fit to past data predict future events?
Swipe through more Machine Learning concepts
Open Pocket Polymath