I want to build a softmax regression model, where the coefficients are assumed to be the same for the different classes, and what varies are the input features. I could only make it work (kind of, still having problems with the intercept) for two classes, using a Logistic Regression model. I am howe…

Apologies, I hit the shape error and assumed that was your problem :blush: You can recover the appropriate behavior by using pm.Multinomial in place of pm.Categorical y_mn = np.array([[0,1] if t == 0 else [1,0] for t in y]) with pm.Model() as m1: b = pm.Normal('b', 0, 20, shape=x1.shape[1]) …

Softmax Regression with shared coefficients between classes

Questions

ricardoV94 September 6, 2019, 9:24am 6

@chartl Thank you so much, that definitely fixed it!

Is there any theoretical reason why pm.Categorical should fail in this case?

Topic		Replies	Views
Implementing Softmax Regresion Questions	3	2262	July 9, 2018
Issues with specification of coefficients in Hierarchichal Softmax Regression Questions	6	848	November 15, 2019
Implementing Logistic Regression with OVR using PyMC3 Questions	2	746	March 27, 2018
Implementing softmax regression with a latent Gaussian variable Questions	4	1864	May 26, 2019
Mixture with multiple observations Questions	6	2347	December 14, 2017

Softmax Regression with shared coefficients between classes

Related topics