N = len(X)
n = MLPRegressor(
hidden_layer_sizes=(10, ),
activation="logistic",
solver="lbfgs",
learning_rate_init=0.1,
batch_size=N,
max_iter=1000,
)
n.fit(X.reshape(-1, 1), y.flatten())
pred = n.predict(X_VAL.reshape(-1, 1))
mse = mean_squared_error(y_val, pred)
plt.plot(X_VAL, pred, label=f"MLP (10 Hidden) ({name}) mse={mse:.6f}")
# RBF
n = RBF_NET(np.arange(0, 2 * np.pi, 2 * np.pi / rbfs_units), rbfs_var)
n.train_batch(X, y)
pred = n.predict(X_VAL)
mse = mean_squared_error(y_val, pred)
plt.plot(
X_VAL,
pred,
label=
f"RBF (units={rbfs_units},var={rbfs_var}) ({name}) mse={mse:.6f}",
)
plt.title("Regressor Performance\n")
plt.legend()
plt.savefig("pictures/3_2_rbf_vs_mlp.png", bbox_inches='tight')
