def solve(models):
k = len(models)
P = np.hstack([m.samples() for m in models])
# P = P.T
# np.random.shuffle(P)
# P = P.T
m, N = P.shape
X_D = sdpclstr.sdp(P, k)
return X_D
def main():
P = np.vstack(
(
scipy.stats.multivariate_normal.rvs(mean=[-1, 2], cov=1, size=20),
scipy.stats.multivariate_normal.rvs(mean=[-1, -1], cov=1, size=20),
scipy.stats.multivariate_normal.rvs(mean=[2, 0], cov=1, size=20),
)
).T
Y = np.array([[-1, -1, 2], [2, -1, 0]])
X_D = sdpclstr.sdp(P, 3)
PX_D = P @ X_D
# fig, ax = pp.subplots(nrows=1, ncols=2)
# fig.set_size_inches(8, 4)
fig1 = pp.figure(figsize=(4, 4))
fig2 = pp.figure(figsize=(4, 4))
# Set axis limits
fig1.gca().set_xlim(-4, 4)
fig2.gca().set_xlim(-4, 4)
fig1.gca().set_ylim(-4, 4)
fig2.gca().set_ylim(-4, 4)
# Set ticks
fig1.gca().set_xticks([-4, -2, 0, 2, 4])
fig1.gca().set_yticks([-4, -2, 0, 2, 4])
fig2.gca().set_xticks([-4, -2, 0, 2, 4])
fig2.gca().set_yticks([-4, -2, 0, 2, 4])
# Plot centers
fig1.gca().plot(Y[0, :], Y[1, :], "xk", ms=15)
fig2.gca().plot(Y[0, :], Y[1, :], "xk", ms=15)
# Plot samples
fig1.gca().plot(P[0, :], P[1, :], "ow")
fig2.gca().plot(PX_D[0, :], PX_D[1, :], "ow")
# fig.show()
fig1.savefig("../figures/denoising-noisy.eps", dpi=160)
fig2.savefig("../figures/denoising-denoised.eps", dpi=160)
