Skip to content

rougier/Review-momentum-illustration

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 

Repository files navigation

README

DOI

This illustration serves to support the argument that structural plasticity may play a crucial functional role. It tests if a delay in the development of the motor pathway could aid learning, by using an analogy with speed and acceleration.

Momentum guided reinforcement learning can be understood using an analogy of velocity and acceleration. In this illustration, velocity corresponds to Hebbian learning and represents the momentum of the trajectory, while acceleration corresponds to reinforcement learning and can modify the velocity vector up to a certain degree (10% in the figure). The illustration simulates the trajectory from a fixed initial position to a target, when the direction of the velocity is continually modified by the acceleration. In the figure, four trajectories with increasing initial delays (left to right) are shown. These delays correspond to the moment when velocity is activated/allowed to initiate the trajectory. Before that, the acceleration influences the direction of the velocity, however without any displacement. When there's no delay, the initial velocity vector can potentially drive the trajectory away from the target (as shown in this specific example), while with longer delays, the acceleration can explore and identify a more conducive direction in order to exhibit a more direct trajectory to the target.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%