Question regarding normalization
venetisgr opened this issue · 4 comments
I would like to apologize in advance if this is the wrong place to ask a question. I have a question regarding the data normalization and the pca feature creation. If your dataset had a shape of (batches,mfcc,samples) how would you apply the transforms I mentioned before. In your code, are you normalizing and extracting pca features for each training example alone ?
Sure, you can ask questions here. I didn't fully understand your question though. Can you elaborate?
Yes of course, when you normalized the samples, did you normalize each sample individually given its current values or did you check all the available samples to get their mean for example and then normalize?
For PCA, the pre-processing step includes StandardScaler which standardizes the features. For the Mel-features, I didn't explicitly normalize standardize the data, but I used BatchNormalization as the initial layer, which kinda does this for me automatically, but based on mini-batch statistics.
feel free to ask any other questions you may have