09:15 < kalingeri> @zoq can you please review the comment on kathirsubavathi initialization tests as soon as possible ?

17:52 < uzipaz> zoq: thank you for all the help yesterday, I was able to successfully run rnn on my dataset with some results

18:03 < zoq> uzipaz: Sure no problem, I think it would be a good idea to use a binary representation for your classes.

18:11 < uzipaz_> zoq: the output class is already binary, I am using BinaryClassificationLayer for the NN output in the code... should I also convert my nominal attributes to binary by adding dummy variables?

18:17 < zoq> uzipaz_: Not exactly, something like 1 0 0 for class one 0 1 0 for class two and 0 0 1 for class three etc.

18:19 < uzipaz_> zoq: I don't understand, the target/output feature is cancer and has only two classes, cancer (1) or no cancer(0)

18:25 < uzipaz_> zoq: I plan to convert all the features (they are all nominal) to binary by adding dummy variables.... is it a good idea to use sigmoid activation function in that case? I think we can just use binary activation function, if x < 0 then f(x) = 0 else f(x) = 1