人工神经网络的源码编程

				/*
				 * public/ann/annfunc/sign.c
				 * 2004-05-03 by ShanLanshan
				 */
				#include "public/ann/annfunc/annfunc.h"
				#include "public/ann/annfunc/sign.h"
				
				namespace ann {
				namespace annfunc {
				namespace sign {
				
					static bool init();
					static void finalize();
					static ann_float main_func(ann_float f);
					static ann_float derivative_func(ann_float f);
					static ann_float derivative_func2(ann_float f);
					static void multi_main_func(ann_float *src, ann_int len, ann_float *dst);
					static void multi_derivative_func(ann_float *src, ann_int len, ann_float *dst);
					static void multi_derivative_func2(ann_float *src, ann_int len, ann_float *dst);
				
					extern  AnnFuncGroup g_funcgroup = {
						"sign",
						"符号函数",
						init,
						finalize,
						main_func,
						derivative_func,
						NULL,
						multi_main_func,
						multi_derivative_func,
						NULL
					};
				
					static bool init()
					{
						return true;
					}
				
					static void finalize() {}
				
					/* 
					 * 符号函数 
					 * w = 1 / ( 1 + e ^ (-f))
					 */
					static ann_float main_func(ann_float f)
					{
						if (f >= 0.0)
							return 1;
						else
							return -1;
					}
				
					/*
					 * 符号函数的导数
					 * 符号函数是不可导的，我们把它当作双曲正切函数看
					 */
					static ann_float derivative_func(ann_float f)
					{
						ann_float u;
						u = pow(E, -f);
						return 2 * u / ((1 + u) * (1 + u));
					}
				
					static void multi_main_func(ann_float *src, ann_int len, ann_float *dst)
					{
						ann_float f;
				
						assert(src != NULL);
						if (dst == NULL)
							dst = src;
				
						for (int i=0; i							f = src[i];
							dst[i] = f >= 0.0 ? 1 : -1;
						}
					}
					
					static void multi_derivative_func(ann_float *src, ann_int len, ann_float *dst)
					{
						ann_float u, f;
				
						assert(src != NULL);
						if (dst == NULL)
							dst = src;
				
						for (int i=0; i							f = src[i];
							u = pow(E, -f);
							dst[i] = 2 * u / ((1 + u) * (1 + u));
						}
					}
				
				}}} // namespace ann::annfunc::sign