Also
Constant Learning — is a term for learning that occurs
continuously (
not merely
continually).
It describes the ability of an adaptive entity to continuously learn as it interacts with its milieu. In other words, a system that does not need to be switched between training, and responding.
Continuous, or Constant Learning is not the same as the term
"on-line learning" as it has been used in the literature. Though often vaguely defined, the term "on-line learning" has come to mean systems that can be switched in and out of a training mode. This allows further training of new responses after the network has been trained on an initial set of exemplars (not allowed in traditional network
learning algorithms). On-line learning networks usually employ special tactics to work around the
catastrophic interference that occurs in traditional
neural networks.
Compare "on-line learning" with constant, or continuous learning, which permits the adaptive system to learn from every interaction with the outside world, directly from the interactions, without switching between response-mode, and training-mode.