The use of "turn" pybrain-an open source Python neural network Toolkit

Original Address http://lavimo.blog.163.com/blog/static/2149411532013911115316263/

Yesterday's main activity is to find a neural network package .... = =
Here, we have to spit out the pybrain before we describe the bag.
First of all, Matlab is the simplest, and very light send you can use a visual tool to learn without brains. However, this is the fool of Matlab, my notebook is 32 bits +2g memory, my input data about 200m, on the memory is not enough, this is obviously a deadly insult to my computer!
Then there are Neurolab, the basic function of this package is the same as MATLAB, but also brought the same problem, still can't bring up, really rubbing force. Also reported section error, this is Python's package ah, you quote a paragraph what is the meaning of the mistake!
Fann,c language written, I want to use the Python interface, I think I can quickly, well, I admit, I am not useless, why, the input data is too big, and there is no Python interface documents! Brother, you are open source software how can this =.
Finally, when I was on the verge of collapse, Pybrain appeared and saved me. In fact, this bag is also very time, normal running is very slow, but can run.

His official homepage is as follows http://www.pybrain.org/docs/, in fact the basic work process on the homepage is very clear.
Here are a few things to illustrate, where you can customize functions between each layer, such as sigmoid or something.

1. Establish a neural network

The first four numbers are the number of layers, you can define unlimited multilayer, but 1-2 hidden layers are recommended.

2. Set up a data set

The dataset is a specific package, 100 is the number of input units, 24 output units, add addsample to increase, there is no other way. It's actually a two-numpy array.

3.
This is training, the construction of the neural network and data structure at the same time input, other parameters can be adjusted.
The epochs of training is the number of iterations

4.

Return the result, also need to run with NN, then return a numpy result.

It seems to be nothing, if you need to continue to refer to the document.

The use of "turn" pybrain-an open source Python neural network Toolkit