Term
Underfitting
Definition
Underfitting is when an AI model is too simplistic and fails to detect underlying patterns in the data it is trained on. This usually results in the model performing poorly when making predictions.
Where you’ll find it
In AI, underfitting can be identified during the model training phase. You won't see it mentioned explicitly in a menu or panel, but it’s recognized by evaluating the model's performance on training data.
Common use cases
- Improving model design by increasing its complexity.
- Adjusting training processes to better capture data nuances.
- Evaluating the sufficiency of the model’s structure during the training phase.
Things to watch out for
- Mistaking underfitting for overfitting, which is its opposite.
- Oversimplifying the model to the extent that it cannot learn effectively.
- Ignoring the importance of selecting the right features and data for training.
Related terms
- Model Complexity
- Feature Selection
- Training Data
- Machine Learning
- Overfitting
Pixelhaze Tip: If you suspect your model is underfitting, try adding more features to your data or increasing the complexity of the model. Simple models might not capture all the nuances needed to make accurate predictions.