Added Examples Folder and MSE Example #158
Conversation
Changed bandstructure to bandstructure_col and dos to dos_col
automatminer/automl/adaptors.py
Outdated
| @@ -118,9 +118,9 @@ def fit(self, df, target, **fit_kwargs): | |||
| self._features = df.drop(columns=target).columns.tolist() | |||
| self._ml_data = {"X": X, "y": y} | |||
| self.fitted_target = target | |||
| self.logger.info("TPOT fitting started.") | |||
There was a problem hiding this comment.
Why are there these logger changes happening here? In this PR we shouldn't do this
automatminer/examples/mse_example.py
Outdated
| @@ -0,0 +1,45 @@ | |||
| import unittest | |||
There was a problem hiding this comment.
Looks good. Also is cool that it's in a test, but I'm afraid it will just confuse people who come to use it (ie, "woah, it is weird they put this test here, I wonder where the example is"). I don't think we will be frequently running this test anyway (especially because you are using the default config, not the debug config which is much faster).
There was a problem hiding this comment.
Instead of being in a test, could you write out the file in a simple script (or even better, notebook), with perhaps a few more comments? Not everyone is particularly familiar with pandas/automatminer/matminer/machine learning, and there are a few areas that won't make sense to someone unfamiliar with this stack.
For example, when we are renaming the formula column, just add a comment saying "The preset automatminer uses pre-defined column names 'composition' and 'structure' to find the composition and structure columns. You can change these by editing your config"
|
@ADA110 thanks for taking care of that issue, Is this ready (or about ready to merge)? |
|
@ardunn If you think everything is good with the iPython file I added, then yes |
The example uses the elastic_tensor_2015 dataset and a default config to create a MatPipe. This MatPipe is used to benchmark the target property K_VRH.
The unit tests confirm that the output of the benchmark is not empty. They also ensure that, based on this specific example, the mean squared error is between 0 and 500.
For debugging purposes, you can use the debug config instead. In addition, make the range of the mean squared error be 0 - 1000 rather than 0 - 500.