« first day (131 days earlier)      last day (2583 days later) » 

Toros91
Toros91
07:00
Hi all,

I'm a beginner in Deep Learning, I'm having a dataset which has consists of encoded data as an input feature and a target variable which states to which class(a total of 12 classes) each record belongs to, multi-class classification problem.

Now, I want to use some deep learning techniques to predict for the new data in python. The reason for me to say that I want to use deep learning as the given data is encoded data.

Any suggestions/pointers on where to start?
 
4 hours later…
Neil Slater
Neil Slater
10:31
@Toros91: Take a look at Keras, and study some of the examples given at keras.io - you can probably adapt the getting started example here keras.io/#getting-started-30-seconds-to-keras as an initial model.
Toros91
Toros91
Thank you very much @Neil Slater, anything else other than keras. I was planning to use Keras and mostly it should be ready by tomorrow.
Neil Slater
Neil Slater
Not really. There is a lot to learn about how all the options and hyper-parameters work. Probably most important is normalise your inputs. Each input should be a fixed length vector, and each element of the vector should have mean 0, standard deviation 1.0 across the dataset. SKlearn's StandardScaler will do that for you.
2
If you have sequential data (e.g. your input is variable length, like text from emails/tweets), then you have more work to do . . . I was assuming not if this is a first exercise
Oh, one more thing, if you don't know already from other ML, remember to set aside some data for validation and testing. Keras will do the validation automatically for you from the training dataset, but usually in addition you want an unbiased test of how good your best solution is once you have found it. You need some test data for that.
Again sklearn can help you there, it has methods to do train/test split
 
2 hours later…
Toros91
Toros91
13:04
sure will start right away. Thank you

« first day (131 days earlier)      last day (2583 days later) »