u1234x1234/kaggle-yelp-restaurant-photo-classification

Winning solution scripts

kaggle-yelp-restaurant-photo-classification's Issues

Problem with Classifier Chain Implementation

Hi,

I am trying to use your classifier chain implementation on my set of features and using an xgboost classifier instance. I have double checked my training feature set and the training labels that I pass to the classifier chain instance. But, the predictions by using classifier chains drastically reduces. If I just use a simple xgboost classifier without chaining them, I get an F1 score of around 0.80 both on training and testing set. But, by using classifier chains, my F1 score goes down to 0.50 which is really low. Any thoughts on why would this be?

Thanks,
Pawan

Clarification of feature extraction and processing

I would be very grateful if you could clarify your feature extraction workflow as much of the file compress.py is commented out. Is the following correct?

1. extract features from the training data using n ImageNet pretrained models

2. normalise each feature matrix separately using sklearn.preprocessing.normalize

3. concatenate all feature matrices horizontally (axis=1)

4. calculate the column mean 'm' (axis=0) and subtract

5. Apply svd using sklearn.decomposition.TruncatedSVD with n_components=64 and algorithm='arpack'

6. repeat for the test data: normalise, concatenate, subtract the mean 'm' (calculated from the training data in step 3), transform using svd (which was fit on the training data in step 4).

Thank you.