Table 3 List of hyperparameters and parameter initializations
From: Geometry of neural dynamics along the cortical attractor landscape reflects changes in attention
Learning rate | 2.5e-5 |
Training iterations | 2500 |
Batch size | 300 |
\({W}_{S}\), \({W}_{1}\), \({W}_{2}\) | \(\begin{array}{c}{{\mathscr{N}}}(0,{0.01}^{2})\end{array}\) |
\(D\) | \(\begin{array}{c}{{\mathscr{N}}}(5,{0.5}^{2})\end{array}\) |
\(\alpha\) | \(\begin{array}{c}{{\mathscr{N}}}(5,{0.05}^{2})\end{array}\) |
\(b\) | 20/3 |
\(\beta\) | \(\begin{array}{c}{{\mathscr{N}}}(0,{0.01}^{2})\end{array}\) |
\({\lambda }_{1}\) | 0.075 |
\({\lambda }_{2}\) | 0.2 |
\({\lambda }_{3}\) | 0.05 |
