Error in Feature Extraction step

[Yash Bolisetty]

Hey Paul,
Amazing notebook about the motor imagery classification. I was trying to replicate this on my machine but I was running into some difficulties in the feature extractions step. I am getting all of the output to all the prior steps (eventhough it is taking quite some time to plot the previous graph), but i am running into the following error:
File "", line 1, in
File "", line 14, in create_fbcsp
File "/Users/yashbolisetty/anaconda3/lib/python3.6/site-packages/sklearn/pipeline.py", line 618, in init
self._validate_transformers()
File "/Users/yashbolisetty/anaconda3/lib/python3.6/site-packages/sklearn/pipeline.py", line 652, in _validate_transformers
self._validate_names(names)
File "/Users/yashbolisetty/anaconda3/lib/python3.6/site-packages/sklearn/pipeline.py", line 75, in _validate_names
'{0!r}'.format(list(names)))
ValueError: Names provided are not unique: ['pipe', 'pipe', 'pipe', 'pipe', 'pipe', 'pipe', 'pipe', 'pipe', 'pipe', 'pipe', 'pipe', 'pipe']
Traceback (most recent call last):
File "", line 1, in
NameError: name 'fbcsp' is not defined
Traceback (most recent call last):
File "", line 1, in
NameError: name 'X_train_feats' is not defined

I have tried to assess this error but I am at a dead end. Any help given would be appreciated.
Thanks you so much once again!!

[Paul Bustios]

Hi @Yash-Bolisetty ,

Thanks for reporting the problem. What version of sklearn are you using? It looks like your sklearn version does not accept repeated names for the transformers.

In the function create_fbcsp, try replacing:

for low in bands:
        pipeline_list.append(
            ("pipe", Pipeline([
                ("filter", BandPassFilter(low, low + band_width)),
                ("csp", CSP(n_components=n_csp_components))
            ]))
        )

by

for low in bands:
        pipeline_list.append(
            Pipeline([
                ("filter", BandPassFilter(low, low + band_width)),
                ("csp", CSP(n_components=n_csp_components))
            ])
        )

or give a different name for each Pipeline appended to pipeline_list.

[Yash Bolisetty]

Hey @bustios thanks for the quick reply. I am currently using sklearn version 0.18.1. I tried replacing the code you showed and was faced with this error:

Traceback (most recent call last):
File "", line 1, in
File "", line 14, in create_fbcsp
File "/Users/yashbolisetty/anaconda3/lib/python3.6/site-packages/sklearn/pipeline.py", line 618, in init
self._validate_transformers()
File "/Users/yashbolisetty/anaconda3/lib/python3.6/site-packages/sklearn/pipeline.py", line 649, in _validate_transformers
names, transformers = zip(*self.transformer_list)
TypeError: zip argument #1 must support iteration
Traceback (most recent call last):
File "", line 1, in
NameError: name 'fbcsp' is not defined
Traceback (most recent call last):
File "", line 1, in
NameError: name 'X_train_feats' is not defined

How should I attempt to fix this or should I just try a different version of sklearn?

[Yash Bolisetty]

@bustios If it would be easier, we could get in a google hangout call to discuss these issues or any further issues?

[Yash Bolisetty]

@bustios I also tried to make the names unique by adding some random number to each of the Pipelines but I got the same error I think:

Traceback (most recent call last):
File "", line 1, in
File "", line 14, in create_fbcsp
File "/Users/yashbolisetty/anaconda3/lib/python3.6/site-packages/sklearn/pipeline.py", line 618, in init
self._validate_transformers()
File "/Users/yashbolisetty/anaconda3/lib/python3.6/site-packages/sklearn/pipeline.py", line 652, in _validate_transformers
self._validate_names(names)
File "/Users/yashbolisetty/anaconda3/lib/python3.6/site-packages/sklearn/pipeline.py", line 75, in _validate_names
'{0!r}'.format(list(names)))
ValueError: Names provided are not unique: ['pipe6', 'pipe1', 'pipe4', 'pipe4', 'pipe6', 'pipe10', 'pipe6', 'pipe4', 'pipe4', 'pipe2', 'pipe7', 'pipe9']
Traceback (most recent call last):
File "", line 1, in
NameError: name 'fbcsp' is not defined
Traceback (most recent call last):
File "", line 1, in
NameError: name 'X_train_feats' is not defined

Once again, if there is a better way for communication and solving these issues, please feel free to mention them.

Thanks so much,

Yash Bolisetty

[Yash Bolisetty]

Hey Paul,
Any help for the errors above are appreciated! Wondering what Sklearn version you used as well?
Thanks,
Yash B

[Paul Bustios]

Hi @Yash-Bolisetty ,
sorry for the late response. I used sklearn version 0.17.1.
When you added random numbers to the names of the pipelines, the resulting names were not unique:

ValueError: Names provided are not unique: ['pipe6', 'pipe1', 'pipe4', 'pipe4', 'pipe6', 'pipe10', 'pipe6', 'pipe4', 'pipe4', 'pipe2', 'pipe7', 'pipe9']

that's why you still have the same error.

[Yash Bolisetty]

Yeah @bustios I had it figured out after looking at it for some time. Thanks anyways for the help!!!!