Problem using MISO

[micheloz]

Applying the example of article "SysIdentPy: A Python package for System Identificationusing NARMAX models" gives an error.
Typing :
model = PolynomialNarmax(non_degree=2,order_selection=True,ylag=2, xlag=[[1, 2], [1, 2]],info_criteria='aic', estimator='least_squares',)
Gives : -

TypeError Traceback (most recent call last)

in ()
----> 1 model = PolynomialNarmax(non_degree=2,order_selection=True,ylag=2, xlag=[[1, 2], [1, 2]],info_criteria='aic', estimator='least_squares',)

2 frames

/content/gdrive/My Drive/python/03_WP/sysidentpy/base.py in (.0)
116 # create only the lags passed from list
117 x_vec_tmp = []
--> 118 x_vec_tmp.extend([lag + 1000*np.ones(np.size(lag)) for lag in xlag])
119 x_vec_tmp = np.array(x_vec_tmp)
120 elif isinstance(xlag, int) and n_inputs == 1:

TypeError: can only concatenate list (not "int") to list

[wilsonrljr]

Hi @micheloz . You just have to set the n_inputs=2. You can check some examples here: http://sysidentpy.org/examples/multiple_inputs_example.html

PolynomialNarmax(non_degree=2, order_selection=True, ylag=2, xlag=[[1, 2], [1, 2]], info_criteria='aic', estimator='least_squares', n_inputs=2)

Let me know if everything works as expected with that modification. There were some modifications and new features since the article was published, so I suggest you to take a look on our documentation.

[micheloz]

Hi, thanks it works fine with following additional options (given in http://sysidentpy.org/examples/multiple_inputs_example.html) :
"n_info_values=15, n_terms=4, extended_least_squares=False"
However, there is still something wrong as the example given in the paper (without these options) leads to above errors.
Regards

[wilsonrljr]

Thanks, @micheloz. The paper is missing the "n_inputs" parameter. As we show in the docs, now you have to set the number of inputs when modeling a multiple input system.

The paper is a reference to cite our package, but since we are implementing changes and new features all the time, the best place to get into SysIdentPy is the package documentation. Actually, we try not to change much of the current state, even if we add new feature. But the "n_inputs" is really missing in the paper.

I'm closing this issue, but feel free to re-open or open a new one if get some error using the package.