Table 1 Simulation parameters
From: A neural implementation model of feedback-based motor learning
Parameter | Definition | Value |
|---|---|---|
dt | Time step | 10 ms |
τ | Time constant | 50 ms |
Δ | Feedback delay | 120 ms |
N | Number of neurons | 400 |
Φ | Nonlinearity | ReLU |
y | Neural activity | – |
s | Stimulus | – |
ϵ | Position error | – |
r | Eligibility trace | – |
W | Recurrent weight matrix | – |
b | Recurrent offset | - |
Win | Input weight matrix | – |
Wout | Output weight matrix | – |
bout | Output offset | – |
F | Feedback weight matrix | – |
v | Velocity (2D) | – |
p | Position (2D) | – |
α | Gradient Descent: Learning rate | 0.001 |
B | Gradient Descent: batch size | 20 |
β | Gradient Descent: weight regularisation | 0.001 |
γ | Gradient Descent: activity regularisation | 0.002 |
η | Feedback-driven plasticity: learning rate | 0.00002 |
T | Reach trajectories: number of time steps in a trial | 300 |
