Ground-truth
In the field of machine learning,the truthRefers to the accurate setting value of the training set for the classification result in supervised learning, which is generally used for error estimation and effect evaluation.
In supervised learning, labeled data usually appears in the form of (x, t), where x represents the input data and t represents the label. The correct label is Ground-Truth, and the wrong label is not (some people also call all labeled data Ground-Truth).
In other words, Ground-truth is the reference standard, which is generally used for error quantification. For example, if you want to predict the temperature of a certain time period based on historical data, then the Ground-truth is the actual temperature of the corresponding time period, and the predicted temperature is the Error.
Ground-truth is also used in the Reinforcement Learning mechanism, which adds a reward and punishment mechanism to learning. For example, the closer the output of the program is to the Ground-truth, the greater the weight of the data that produces this result.