inductive bias
Inductive bias is the set of assumptions a learning algorithm uses to generalize beyond the data it was trained on.
Inductive bias refers to the assumptions or preferences a machine learning model or learning algorithm relies on to generalize from finite training data to data it has not seen. It is embedded in choices such as model architecture, training methods and regularization.
A convolutional neural network, for example, builds in the assumptions that nearby pixels are closely related and that the same pattern can appear anywhere in an image. A well-chosen inductive bias helps a model learn from less data, while a wrong or overly strong one can limit performance.
This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.