UNDERFITTING VS OVERFITTING
| No | Feature | Underfitting | Overfitting |
|---|---|---|---|
| 1 | Definition | Model learns too little from data | Model learns too much from data |
| 2 | Performance | Poor on training and testing data | Good on training but poor on testing data |
| 3 | Cause | Model too simple | Model too complex |
| 4 | Learning Ability | Cannot capture patterns | Memorizes data instead of learning |
| 5 | Accuracy | Low accuracy | High training accuracy, low test accuracy |
| 6 | Error Rate | High error on both datasets | Low training error, high testing error |
| 7 | Generalization | Poor | Poor |
| 8 | Example | Predicting marks with very basic rule | Memorizing exact answers of training data |
| 9 | Model Behavior | Misses important relationships | Too sensitive to training data |
| 10 | Solution | Increase model complexity | Reduce complexity, use regularization |
SIMPLE EXPLANATION
1 Underfitting
A Model is too simple
B Does not learn properly
C Gives wrong predictions
2 Overfitting
A Model is too complex
B Learns too much detail
C Fails on new data
REAL LIFE EXAMPLE
1 Underfitting → Student studies very little and fails
2 Overfitting → Student memorizes answers but cannot handle new questions
CONCLUSION
Underfitting and overfitting are common problems in Machine Learning. A good model should balance learning so it performs well on both training and testing data. Proper tuning and practice help achieve better results.
0 Comments