Let's discuss video "Machine Learning Theory - Underfitting vs Overfitting"
Of course one should not overfit or underfit. This is a common problem which also occures in the regression model. There it is simple to find out if the parameters lead to an overfitted or an underfitted plot. It's overfitted if the standard deviation of the bands is low and the standard deviation of the slope is high, and underfitted if the standard deviation of the bands is high and the standard deviation of the slope is low.
In machine learning it is comparable. You have the bias and the variance and they need to be in balance. Unfortunately this video doesn't explain how to do that, it just explain how important it is.
Note: Depending on the used model, the underfitted values can do well, but the perfectly fitted values will do even better. This is one case why machine learning should not be overrated, but should be used on top of good and meaningful datas, which come from good indicator-types / -models.
The tutorial from the og himself
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
What is Underfitting?
Oversimplifying the problem
Does not do well in the training set
Error due to bias