What is biased sampling in machine learning, and how can it affect model training?

Biased sampling occurs when certain members of a population are more likely to be selected for the training dataset than others, leading to an unrepresentative sample. Such sampling can result in overfitting, where the model learns patterns specific to the biased sample rather than generalizable patterns across the entire population. This can degrade the model's performance on unseen data, leading to inaccurate predictions. To mitigate biased sampling, techniques such as stratified sampling, random sampling, or ensuring diversity in the dataset can be used.