From d3e24f3cbff8a5f1503ff45cdf8d1d8118e4c049 Mon Sep 17 00:00:00 2001
From: Nghia Ho <nghiaho12@yahoo.com>
Date: Fri, 3 Jan 2014 14:18:07 +1100
Subject: [PATCH] Improved documentation for neural network

---
 modules/ml/doc/neural_networks.rst | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/modules/ml/doc/neural_networks.rst b/modules/ml/doc/neural_networks.rst
index 0496e2201a..776bf243bd 100644
--- a/modules/ml/doc/neural_networks.rst
+++ b/modules/ml/doc/neural_networks.rst
@@ -240,6 +240,7 @@ This method applies the specified training algorithm to computing/adjusting the
 
 The RPROP training algorithm is parallelized with the TBB library.
 
+If you are using the default ``cvANN_MLP::SIGMOID_SYM`` activation function then the output should be in the range [-1,1], instead of [0,1], for optimal results.
 
 CvANN_MLP::predict
 ------------------
@@ -257,6 +258,8 @@ Predicts responses for input samples.
 
 The method returns a dummy value which should be ignored.
 
+If you are using the default ``cvANN_MLP::SIGMOID_SYM`` activation function with the default parameter values fparam1=0 and fparam2=0 then the function used is y = 1.7159*tanh(2/3 * x), so the output will range from [-1.7159, 1.7159], instead of [0,1].
+
 CvANN_MLP::get_layer_count
 --------------------------
 Returns the number of layers in the MLP.