NormalMixture regression

Questions
5

In a toy example, I’m able to use the tile function to produce a working example:

import scipy
import theano
import pymc3 as pm

X = np.array([[1, 0], [1, 0], [0, 1], [1, 0], [1, 0], [0, 1], [0, 1], [0, 1], [0, 1], [1, 0], [0, 1], [0, 1], [1, 0]])

X = scipy.sparse.csr_matrix(X)
Y = np.array([[5.28708186], [4.96509597], [0.96509597], [ 5.3526662], [6.19951688], [-0.96509597], [0.49597], [-0.56509597], [0.00053345], [4.19951688], [-10], [-10], [-10]])

X = theano.sparse.as_sparse_variable(X)
Y = theano.tensor.as_tensor_variable(Y)

with pm.Model() as model:
    w = pm.Dirichlet('w', a=np.array([1, 1]))

    sd = pm.HalfNormal('sd', sd=1.0, shape=(1, 1))
    mu = pm.Normal('mu', mu=0, sd=sd, shape=(2, 1))

    mu = theano.sparse.dot(X, mu)
    kappa = pm.HalfNormal('kappa', sd=1.0, shape=(1, 1))

    mu2 = theano.tensor.as_tensor_variable(-10)
    mu2 = theano.tensor.shape_padleft(mu2, 2)
    mu2 = theano.tensor.tile(mu2, reps=[13, 1])

    sd2 = theano.tensor.as_tensor_variable(1e-5)
    sd2 = theano.tensor.shape_padleft(sd2, 2)

    pm.NormalMixture('likelihood', w=w, mu=theano.tensor.stack([mu, mu2]), 
                     sd=theano.tensor.stack([kappa, sd2]), observed=Y)

However, when I use this methodology in my current model factory function, I get the following traceback.

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-149-56c6fce440e4> in <module>
      2 preprocessors['design_matrix'].fit(train_data)
      3 
----> 4 model = construct_model(train_data, preprocessors, fit_intercept=False)
      5 
      6 with model:

<ipython-input-148-0dcee7f04157> in construct_model(data, preprocessors, fit_intercept)
     47 
     48         pm.NormalMixture('likelihood',  w=w, 
---> 49                          mu=theano.tensor.stack([mu, degenerate_mu]),
     50                          sd=theano.tensor.stack([kappa, degenerate_sd]),
     51                          observed=success_rate)

/anaconda3/envs/audience-concentration-forecast-service-env/lib/python3.6/site-packages/theano/tensor/basic.py in stack(*tensors, **kwargs)
   4726         dtype = scal.upcast(*[i.dtype for i in tensors])
   4727         return theano.tensor.opt.MakeVector(dtype)(*tensors)
-> 4728     return join(axis, *[shape_padaxis(t, axis) for t in tensors])
   4729 
   4730 

/anaconda3/envs/audience-concentration-forecast-service-env/lib/python3.6/site-packages/theano/tensor/basic.py in join(axis, *tensors_list)
   4500         return tensors_list[0]
   4501     else:
-> 4502         return join_(axis, *tensors_list)
   4503 
   4504 

/anaconda3/envs/audience-concentration-forecast-service-env/lib/python3.6/site-packages/theano/gof/op.py in __call__(self, *inputs, **kwargs)
    613         """
    614         return_list = kwargs.pop('return_list', False)
--> 615         node = self.make_node(*inputs, **kwargs)
    616 
    617         if config.compute_test_value != 'off':

/anaconda3/envs/audience-concentration-forecast-service-env/lib/python3.6/site-packages/theano/tensor/basic.py in make_node(self, *axis_and_tensors)
   4233 
   4234         return self._make_node_internal(
-> 4235             axis, tensors, as_tensor_variable_args, output_maker)
   4236 
   4237     def _make_node_internal(self, axis, tensors,

/anaconda3/envs/audience-concentration-forecast-service-env/lib/python3.6/site-packages/theano/tensor/basic.py in _make_node_internal(self, axis, tensors, as_tensor_variable_args, output_maker)
   4285                             continue
   4286                         if bflag:
-> 4287                             bcastable[current_axis] = True
   4288                 try:
   4289                     bcastable[axis] = False

IndexError: list assignment index out of range

Here is the model factor code snippet.

with pm.Model() as model:
    coefficient_vector = dfn.assemble_coefficient_vector(feature_shapes, fit_intercept=fit_intercept)
    mu = theano.sparse.dot(X, coefficient_vector)
    kappa = pm.HalfNormal('κ', sd=1.0, shape=(1, 1))

    w = pm.Dirichlet('w', a=np.array([1, 1]))

    degenerate_mu = theano.tensor.as_tensor_variable(DEGENERATE_VALUE)
    degenerate_mu = theano.tensor.shape_padleft(degenerate_mu, 2)
    degenerate_mu = theano.tensor.tile(degenerate_mu, reps=[M, 1])

    degenerate_sd = theano.tensor.as_tensor_variable(1e-5)
    degenerate_sd = theano.tensor.shape_padleft(degenerate_sd, 2)

    pm.NormalMixture('likelihood',  w=w, 
                     mu=theano.tensor.stack([mu, degenerate_mu]), 
                     sd=theano.tensor.stack([kappa, degenerate_sd]), 
                     observed=success_rate)

Does this error make sense?