Help me here, /sci/
I have 10k of database entries, with 20 integer atributes. I must classify each entry on one of n categories, based on their values.
Example: Entry A =[8,10,12,8] - Group x
Entry B =[5,0,20,25] - Group y
And so on...
I have the training values for each category, but I can't decide which technique to use - pretty newbie on this subject. Any sugestions?
>only 10k entries
Just try everything you can think of.
Reduce the number of cross-validation folds for those that take too long.
just use sort algorithms, many are already embeded in prog. languages.
if you have them in array of arrays form, just sort by the place or info in subarray (e.g. A[2]=12, following your example)
if you have them in struct form, you have already sorted atributes, just use sort and specify which atrivube in the sorting function
>>7639639
are you retarded
Use a deep restricted Boltzmann machine with N output nodes, and 4 input nodes (since input is multinomial, you'll have to softmax it first). There are plenty of resources on how to do this.
>>7639639
Wow what a retard
>>7639653
By the way, I suggested Boltzmann machine over other classifiers because your outputs are Bernoulli.
>>7639350
standard backprop neural net should do the trick depending on how many training values you have.
>>7639350
Just use k-means
>>7639675
You retarded? He has labeled data.
>>7639657
I'll take a look. Thanks anon!
>>7639658
About 40-50 entries for each category.
>>7639759
Yes, I know, pretty much nothing :(
