Documentation re-generated, a lot of minor stuff

This commit is contained in:
blacklight 2009-08-16 20:57:15 +02:00
parent d52976e74e
commit 7861e56f35
144 changed files with 2589 additions and 851 deletions
examples

View file

@ -1,10 +1,19 @@
This directory contains some sources to illustrate LibNeural++ usage.
learnAdd is a source training a neural network to do simple sums
between two integer positive numbers. The network is then saved to a
binary file called "adder.net". This file can then be used by "doAdd".
In this program you're asked input numbers to the network, and then
the output is given. If executable files would not be present, type
`make` to create them.
* @example examples/learnAdd.cpp Show how to train a network that performs sums between
* two real numbers. The training XML is built from scratch, then saved to a file, then
* the network is initialized using that XML file, trained, and the resulting trained
* network is saved to adder.net. Then, you should take a look at doAdd.cpp to see how
* to use that file to use the network.
Of course, you must have Neural++ already installed before typing `make`
* @example examples/doAdd.cpp Show how to use a network already trained and saved to a
* binary file. In this case, a network trained to simply perform sums between two real
* numbers, that should have already been created using learnAdd.
* @example examples/adderFromScratch.cpp Similar to learnAdd.cpp, but this time the
* training XML is generated as a string and not saved to a file, and parsed by the
* program itself to build the network. Then, the program asks two real numbers, and
* performs both the sum and the difference between them, putting the sum's output on
* the first output neuron and the difference's on the second output neuron. Anyway,
* using more than one neuron in the output layer is strongly discouraged, as the network
* usually won't set correctly the synaptical weights to give satisfying and accurate
* answers for all of the operations.