The Binding of Learning to Action in Motor Adaptation Text S1 Example of credit assignment for error-based learning: the gradient-descent learning rule

Example of credit assignment for error-based learning: the gradient-descent learning rule The familiar gradient-descent learning rule [1-2] provides an example of credit assignment during errorbased learning. This rule updates the state of a learning system so that for a given (arbitrarily small) step size, maximal reduction of error will be achieved for a subsequent trial with the same input. If the state of the learning system can be characterized by a vector of parameters (w), then the output of the system (y) is related to its input (x) by: