# Positions
x = np.arange(len(src)).reshape(-1, 1) # shape: [F, 1]
n_bins = len(src)
# Cost matrix: squared Euclidean distance
C = ot.utils.dist(x, x, metric='sqeuclidean')
# Input distributions
a = src.copy() # shape: (100,)
b = tgt.copy() # shape: (100,)
# Stack them into a matrix: [n_bins, n_distributions]
distributions = np.stack([a, b], axis=1) # shape: [F, 2]
# Weights for interpolation (e.g. t=0.3 → 30% a + 70% b)
t_values = np.linspace(0.1, 0.9, 5)
colors = ['g', 'c', 'm', 'y', 'k']
# Plot original
plt.figure(figsize=(10, 6))
plt.plot(x, 10 * np.log10(a), 'b', label='Source distribution')
plt.plot(x, 10 * np.log10(b), 'r', label='Target distribution')
for t, color in zip(t_values, colors):
 weights = np.array([1 - t, t])
 # Compute unbalanced barycenter
 barycenter = ot.unbalanced.barycenter_unbalanced(
 distributions, C, reg=1e-1, reg_m=10, weights=weights
 ) # shape: [F,]
 plt.plot(x, 10 * np.log10(barycenter), color=color, label=f"t={t:.1f}")
plt.title("Unbalanced OT Barycentric Interpolation (POT)")
plt.legend()
plt.grid(True)
plt.tight_layout()
plt.show()