To understand a little bit more about overfitting and underfitting, I found this example:
Let’s Imagine we have three students, let’s call them student A, B and C respectively
- Student A is not interested in learning and he/she doesn’t pay much attention
- Student B pays attetion in class, but he/she memorizes absolutely all that professor says, not learning the main concepts
- Student C pays attention in class and he/she learns the gist concepts from the classes, not memorizing but learning
Now, the professor decides to take a class test previosly before the official test, the class test would be the training set and the official test would be the test set
|
- Student A obtains 50% in this class test because he/she is not interested in learning and hence in the official test she/he will obtain a bad result as well (Underfitting)
- Student B obtains 98% in this class test because he has memorized all, but he will obtain a lower score in the official test because he/she will have difficulties to face the new information that he/she didn’t see in classes and the class test, since he/she doesn’t learn just memorize (Overfitting)
- Student C obtains quite similiar results in both cases, because this student learned the main concepts and doesn’t have problems with new information (Ideal Case)
|
Reference link: https://www.analyticsvidhya.com/blog/2020/02/underfitting-overfitting-best-fitting-machine-learning/
¿Quieres ver más aportes, preguntas y respuestas de la comunidad?