Talk:Structural learning

The current version provides a well-written and concise definition of structural learning in motor control. The actual motor control section of the entry is currently rather brief (just as long as the cognitive science section, which is not really the main topic of the article). I am sure the motor control section will grow with future editions, but a few additions would already improve the article further, I think.

a) It would be nice to provide in illustrative Figure of the parameter subspace for the visual rotation example.

b) A second experimental example of structural learning in motor control (e.g., force field learning, Kobak & Mehring, 2012) would be instructive.

c) A brief discussion of how to distinguish structural learning from savings in relearning of certain places in the parameter space (independent of learning a full structure) would be informative. It would serve to illustrate the (experimentally often relatively subtle) distinction between parametric and structural learning in a more concrete example.

Minor comments:

Definition: “Such invariants can then be exploited for faster generalization and learning-to-learn…” Shouldn’t this be “faster generalization and learning”? It’s the processes of getting faster at learning that usually is called “learning to learn”.

Adaptive Control: “their range of potential values and the noise properties”. I believe there should be a comma before the “and”.

The in-text formulas do no display on my webbrowser (safari), but are displayed as $formula$.

Bayesian Networks: It is unclear whether you want to say that “later stages of neural processing” constitute “sensory input”. I think under the normal definition of the term they do not.