Machine Learning
Underfitting
Underfitting occurs when a machine learning model is too simple to capture the underlying patterns in the training data. This results in poor performance on both the training data and unseen data, indicating a high bias and low variance.
Explanation
Underfitting arises when a model lacks the capacity to learn the complexities present in the data. This often happens when using a linear model to fit non-linear data, or when the model is not trained for a sufficient amount of time. The model makes strong assumptions about the data that are not accurate, leading to systematic errors. Indicators of underfitting include high training error and high test error, both significantly worse than expected. Addressing underfitting typically involves increasing model complexity (e.g., adding more layers or parameters to a neural network, using a non-linear model), engineering more relevant features from the data, or training the model for a longer duration. Regularization techniques are generally *not* helpful for underfitting, as they penalize complexity, and the model needs *more* complexity to properly fit the data.