Neural ODE

Continuous-depth network as a learned vector field — trajectory integration and phase portrait

Phase Portrait

Trajectories Start End
dz/dt = f_θ(z,t) = W₂·tanh(W₁·z + b₁) + b₂

Vector Field Controls

Preset:
W₁ scale1.0
W₂ scale1.0
Bias b₁0.0
T (depth)3.0
dt step0.05

Loss Curve (spiral to center)