Deep Learning/Neural Nets
islem_h
New Altair Community Member
Hi there,
I am trying to solve a regression problem with RapidMiner.
My question is what is the difference between the included (H2O) operator Deep Learning and the operator neural nets? I see that in both we can add as many hidden layers as we wish. In which case would Deep Learning not needed/appropriate to use?
(In general, can a neural network with 2 hidden layers be called deep neural network?)
Cheers!
Thank you
I am trying to solve a regression problem with RapidMiner.
My question is what is the difference between the included (H2O) operator Deep Learning and the operator neural nets? I see that in both we can add as many hidden layers as we wish. In which case would Deep Learning not needed/appropriate to use?
(In general, can a neural network with 2 hidden layers be called deep neural network?)
Cheers!
Thank you
Tagged:
0
Best Answer
-
The simple Deep Learning operator gives you different activation functions and the ability to add dropout layers. There are also some advanced parameters that allow you to fine tune the algorithm performance which you can read about in the H2O documentation online. So it is somewhat more complex than the basic neural network operator.
However, if you want to get more complex in deep learning model design, you should check out the new (free) Deep Learning extension, which adds a lot more capabilities to creating deep learning models in RapidMiner (at the cost of higher complexity and more setup for the user).
1
Answers
-
Hi @islem_h
Neural networks with more layers are called MLP(Multi-layer perceptrons) which is a subset of deep networks. One major difference is the activation functions which you can set in Deep learning operator it is also one of the significance of deep learning. Please go through below link for a more clear explanation.
https://stats.stackexchange.com/questions/315402/multi-layer-perceptron-vs-deep-neural-network
Please correct me if there is any misconception.
1 -
The simple Deep Learning operator gives you different activation functions and the ability to add dropout layers. There are also some advanced parameters that allow you to fine tune the algorithm performance which you can read about in the H2O documentation online. So it is somewhat more complex than the basic neural network operator.
However, if you want to get more complex in deep learning model design, you should check out the new (free) Deep Learning extension, which adds a lot more capabilities to creating deep learning models in RapidMiner (at the cost of higher complexity and more setup for the user).
1