Here I have skipped the technical stuff and just try to relate model with a human(movie character)that’s it!
If humans are Ml models
Let’s take a scenario first,
Farhan, Silencer & Rancho are three friends who are doing engineering.
- Farhan studies little. He’s is not serious about engineering and his knowledge is poor in this field.
- Silencer studies hard but just memorize the topics. He has not well enough understanding by heart.
- And here the Rancho is passionate about engineering. He understands the topics by heart not only to cut a good result in an exam.
Now, it’s exam time:
1.Farhan just pass the exam.
2.Silencer became second in the exam.
note: Though Silencer cut a good figure in the exam, his solution for unseen problem or data is terrible in real-world cases!
3.Rancho became first in the exam.
And in professional life:
all of them were given a specific problem to solve and the problem is slightly different from academic study.
- The solution that Silencer will provide would be terrible in practical but people may have high faith in him as he has good academic results.
And this is the most dangerous thing! - Farhan’s solution may dum as well but everybody knows his knowledge is poor so people will be aware of it.
- And finally, Rancho will give the best & most appropriate solution.
Now, let’s connect all the things:
- Imagine all the three characters are your Machine Learning models.
- Academic study is the training of the model.
- The exam’s result is the model’s accuracy.
- The problem in professional life is like unseen data.
Here,
Farhan ===> underfitted model.
Silencer===>overfitted model.
Rancho====>wellfitted model.
ML is Fun🎈
Happy Learning🎈