Example identification_of_an_electromechanical_system.ipynb not working

[ghost]

Hello,

First of all, thanks for the creation of this project, tools for working with NARMAX are not easily available. I am currently evaluating SysIdentPy to be used at work and I am going through some examples and found this example throughs up the following error:

ValueError                                Traceback (most recent call last)
Input In [8], in <cell line: 1>()
----> 1 model.fit(X=x_train, y=y_train)
      2 yhat = model.predict(X=x_valid, y=y_valid)
      3 rrse = root_relative_squared_error(y_valid, yhat)

File ~/sysidentpy/sysidentpy/model_structure_selection/forward_regression_orthogonal_least_squares.py:476, in FROLS.fit(self, X, y)
    471 self.regressor_code = self.regressor_space(
    472     self.non_degree, self.xlag, self.ylag, self._n_inputs, self.model_type
    473 )
    475 if self._order_selection is True:
--> 476     self.info_values = self.information_criterion(reg_matrix, y)
    478 if self.n_terms is None and self._order_selection is True:
    479     model_length = np.where(self.info_values == np.amin(self.info_values))

File ~/sysidentpy/sysidentpy/model_structure_selection/forward_regression_orthogonal_least_squares.py:359, in FROLS.information_criterion(self, X_base, y)
    356 n_theta = i + 1
    357 regressor_matrix = self.error_reduction_ratio(X_base, y, n_theta)[2]
--> 359 tmp_theta = getattr(self, self.estimator)(regressor_matrix, y)
    361 tmp_yhat = np.dot(regressor_matrix, tmp_theta)
    362 tmp_residual = y[self.max_lag :] - tmp_yhat

File ~/sysidentpy/sysidentpy/parameter_estimation/estimators.py:255, in Estimators.recursive_least_squares(self, psi, y)
    253 k = np.divide(k_numerator, k_denominator)
    254 self.xi[i, 0] = y[i, 0] - np.dot(psi_tmp.T, theta[:, i - 1])
--> 255 theta[:, i] = list(theta[:, i - 1].reshape(-1, 1) + k.dot(self.xi[i, 0]))
    257 p1 = p.dot(psi[i, :].reshape(-1, 1)).dot(psi[i, :].reshape(-1, 1).T).dot(p)
    258 p2 = (
    259     psi[i, :].reshape(-1, 1).T.dot(p).dot(psi[i, :].reshape(-1, 1))
    260     + self._lam
    261 )

ValueError: setting an array element with a sequence. The requested array would exceed the maximum number of dimension of 1.

Could you please give me a clue of how to make it work?

Have python 3.9, SysIdentPy 0.2.0 running on Mac Monterrey

Thanks in advance

[wilsonrljr]

Hey ghost! Thanks for reaching out. I'll try to reproduce the error later today. Could you try to run that example using the v0.2.1 so we can check if the error is fixed in the last version?

[wilsonrljr]

Hey! No problem regarding your username. I'll investigate this issue this weekend. I tried to replicate it, but I didnt get any error. I'll try again this weekend so I can add any fix in the next update (which is almost done)

[gamcorn]

Thanks! if you need any detail from my runs do not hesitate to ask, I will be glad to help !

[wilsonrljr]

Hey! I tried different thing here and I got the same error only in a specific case: using numpy 1.24.*. Testing with numpy version 1.23.5 or older worked as expected, however the error appeared using the newest version of numpy.

You said you tested older versions of numpy and the error persisted. I tested with several different versions as well and the only one with that problem was the 1.24.*. Could you create a new environment with numpy 1.23.5 and test it again, please? In the new environment, check if the version is right in the kernel after installing SysIdentPy just to make sure.

I'll update the requirements in the next version to not use the 1.24* version of numpy. After that I'll upload another version just to fix this issue with the new versions of numpy.

[gamcorn]

Yes ! that is correct, version 1.24.x is the one that fails. I did something wrong the last time I checked previous numpy versions and jupyter notebook was not refreshing. So the scenario you describe is correct only version 1.24x has the issue. Would you like me to make a pull request with the changes I made for it to work on above 1.24 ?

[wilsonrljr]

Hey @gamcorn , I'll add your fix in the next update. Thank you!