I have a pandas dataframe that describes some fields of the register. I have used one hot encoding to encode the feature vectors that are not numbers. Finally my dataset now has 4000 rows * 4 columns. It contains only numbers. I want to generate the same input using AutoEncoders but I didn't find any useful link that I can use for that. The ones that I used had some dimension problems when I use my data. Does Anyone recommend any helpful tutorial ?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Concerning Encoding, this link helped me a lot. If you try the code in part 'One Hot Encode with scikit-learn', you will get your encoded vectors. You just have to feed it a list of all your tokens. So I have extracted the fields of the register into a list of markups.
As an output of the one hot encoding part, you will get a n dimensional array, you feed it to the script in part 'Let's build the simplest possible autoencoder' of the AutoEncoder link
In the Input vector you need to put how many classes you have (instead of the value 784) and in our case, you specify the number of columns of the nd array which represent the unique tokens.
Then to compare the predicted output, you have to use decoding in order to compare visually the text that was regenerated.