# Neural network to extract correlated columns

I want to use a neural network to find correlated columns in a .csv file and give them as a output. The input .csv file has multiple columns with 0 and 1 ( like Booleans) in it. The file got the assignment from people to interests in it.

Example .csv input:

UserID   History  Math  Physics  Art  Music  ...
User1    0        1     1        0    0      ...
User2    0        0     0        1    1      ...
User3    0        1     1        1    1      ...
User4    1        0     1        1    0      ...
...


The output should be in this case something like: {math,physics}, {art,music}, {history,physics,art} - I exclude here {math,physics,art,music} because in a step afterwards i want to exclude (at least some) which can be created through the combination of others.

At the moment I have a problem that i don´t know which type of neural network could complete this task. How can I solve this problem?

So the important thing, that a column can have more than one column it correlates to - so its not like simple k-means clustering (as far as I understand it).