Top Page | Upper Page | Contents | About This Site | JAPANESE

Deep Learning

Deep learning is one of the application of Neural Network . Difference bettween deep learning and old neural network is the number intermediate layers between Y and X.

Idea to Build Deep Learning

Deep learning includes the many intermediate layers.

There was the problem to calculate many intermediate layers. But deep learning solved the problem by pre-learning.

At Principal component regression analysis, principal component analysis is used before multi-regression analysis . And there are approaches that we use SOM or cluster analysis before decision tree. These methods are used to make clear the conclusion of the analysis and to strengthen the week-point of multi-regression analysis and decision tree.

The idea of pre-learning in deep learning is similar to these approaches.

Free Software

Many package are developed in python. And H2O, TensorFlow, Keras, MXNet, RSNNS and rnn are also used in R. .

NEXT Auto Encoder