Add GaussianMixture #169
Add GaussianMixture #169
Conversation
| n_features = X.type.shape[1] | ||
| n_components = op.means_.shape[0] | ||
|
|
||
| # def _estimate_weighted_log_prob(self, X): |
There was a problem hiding this comment.
To remember where I found the implementation in scikit-learn.
There was a problem hiding this comment.
I think it would be better to have comments instead. That would make it clear to anyone reading the code.
There was a problem hiding this comment.
I think I did (line 29). Does it need more?
| # self._estimate_log_prob(X) + self._estimate_log_weights() | ||
| log_weights = np.log(op.weights_) # self._estimate_log_weights() | ||
|
|
||
| # self._estimate_log_prob(X) |
There was a problem hiding this comment.
Commented code again?
| op.precisions_cholesky_, op.covariance_type, n_features) | ||
|
|
||
| if op.covariance_type == 'full': | ||
| # shape(op.means_) = (n_components, n_features) |
There was a problem hiding this comment.
I see a lot of commented code in this file, could you clean it up?
There was a problem hiding this comment.
I prefer to let it, it is how it is implemented in scikit-learn, I can add a new comment to specify it comes from sklearn.
| def calculate_gaussian_mixture_output_shapes(operator): | ||
| check_input_and_output_numbers(operator, input_count_range=1, | ||
| output_count_range=2) | ||
| check_input_and_output_types(operator, good_input_types=[FloatTensorType]) |
There was a problem hiding this comment.
Why is int not allowed as an input type? Scikit allows int features.
There was a problem hiding this comment.
I hesitate. Statistically, it makes no sense to fix a gaussian mixture on integer data as it cannot be gaussian. I'll fix it.
No description provided.