1、神經(jīng)網(wǎng)絡(luò)算法神經(jīng)網(wǎng)絡(luò)算法第1頁(yè)目標(biāo)內(nèi)容學(xué)習(xí)神經(jīng)網(wǎng)絡(luò)基礎(chǔ)原理與方法。1、人工神經(jīng)元數(shù)學(xué)模型4、BP神經(jīng)網(wǎng)絡(luò)應(yīng)用3、BP神經(jīng)網(wǎng)絡(luò)Matlab工具箱函數(shù) 2、BP神經(jīng)網(wǎng)絡(luò)神經(jīng)網(wǎng)絡(luò)算法第2頁(yè)神經(jīng)網(wǎng)絡(luò)算法第3頁(yè)一、人工神經(jīng)元數(shù)學(xué)模型神經(jīng)網(wǎng)絡(luò)算法第4頁(yè)神經(jīng)網(wǎng)絡(luò)算法第5頁(yè)神經(jīng)網(wǎng)絡(luò)算法第6頁(yè)神經(jīng)網(wǎng)絡(luò)算法第7頁(yè)BP神經(jīng)網(wǎng)絡(luò)拓?fù)浣Y(jié)構(gòu)如圖所表示。1.BP神經(jīng)網(wǎng)絡(luò)結(jié)構(gòu)：神經(jīng)網(wǎng)絡(luò)算法第8頁(yè)神經(jīng)網(wǎng)絡(luò)算法第9頁(yè)2BP神經(jīng)網(wǎng)絡(luò)學(xué)習(xí)算法及流程以三層BP神經(jīng)網(wǎng)絡(luò)為例，它訓(xùn)練過(guò)程包含以下幾個(gè)步驟：神經(jīng)網(wǎng)絡(luò)算法第10頁(yè)神經(jīng)網(wǎng)絡(luò)算法第11頁(yè)神經(jīng)網(wǎng)絡(luò)算法第12頁(yè)BP神經(jīng)網(wǎng)絡(luò)流程圖： 神經(jīng)網(wǎng)絡(luò)算法第13頁(yè)三、BP神經(jīng)網(wǎng)絡(luò)Matlab工具箱

2、函數(shù) 神經(jīng)網(wǎng)絡(luò)算法第14頁(yè)net=newff(PR, S1,S2,SN, TF1,TF2,TFN, BTF, BLF, PF)神經(jīng)網(wǎng)絡(luò)算法第15頁(yè)神經(jīng)網(wǎng)絡(luò)算法第16頁(yè)神經(jīng)網(wǎng)絡(luò)算法第17頁(yè)神經(jīng)網(wǎng)絡(luò)算法第18頁(yè)神經(jīng)網(wǎng)絡(luò)算法第19頁(yè)神經(jīng)網(wǎng)絡(luò)算法第20頁(yè)神經(jīng)網(wǎng)絡(luò)算法第21頁(yè) 網(wǎng)絡(luò)經(jīng)過(guò)177次訓(xùn)練后，即使網(wǎng)絡(luò)性能還沒(méi)有到達(dá)0，不過(guò)輸出均方誤差已經(jīng)很小了，MSE2.95307e-006，誤差曲線如圖1所表示。為更直觀地了解網(wǎng)絡(luò)輸出與目標(biāo)向量之間關(guān)系，見(jiàn)圖2所表示。 plot(P,T,-,P,Y,o)圖1 BP神經(jīng)網(wǎng)絡(luò)訓(xùn)練誤差曲線圖 圖2 訓(xùn)練后BP神經(jīng)網(wǎng)絡(luò)仿真圖To Matlab exp12_4_1.m神

3、經(jīng)網(wǎng)絡(luò)算法第22頁(yè)神經(jīng)網(wǎng)絡(luò)算法第23頁(yè)神經(jīng)網(wǎng)絡(luò)算法第24頁(yè)神經(jīng)網(wǎng)絡(luò)算法第25頁(yè)神經(jīng)網(wǎng)絡(luò)算法第26頁(yè)神經(jīng)網(wǎng)絡(luò)算法第27頁(yè)神經(jīng)網(wǎng)絡(luò)算法第28頁(yè)BP神經(jīng)網(wǎng)絡(luò)測(cè)試結(jié)果圖 To Matlab exp12_4_2.m神經(jīng)網(wǎng)絡(luò)算法第29頁(yè)神經(jīng)網(wǎng)絡(luò)算法第30頁(yè)神經(jīng)網(wǎng)絡(luò)算法第31頁(yè)神經(jīng)網(wǎng)絡(luò)算法第32頁(yè)神經(jīng)網(wǎng)絡(luò)算法第33頁(yè)神經(jīng)網(wǎng)絡(luò)算法第34頁(yè)神經(jīng)網(wǎng)絡(luò)算法第35頁(yè)神經(jīng)網(wǎng)絡(luò)算法第36頁(yè)神經(jīng)網(wǎng)絡(luò)算法第37頁(yè)神經(jīng)網(wǎng)絡(luò)算法第38頁(yè)To Matlab exp12_4_3.m神經(jīng)網(wǎng)絡(luò)算法第39頁(yè)神經(jīng)網(wǎng)絡(luò)算法第40頁(yè)練習(xí)1、蠓蟲(chóng)分類問(wèn)題 生物學(xué)家試圖對(duì)兩種蠓蟲(chóng)（Af與Apf）進(jìn)行判別，依據(jù)資料是觸角和翅膀長(zhǎng)度，已經(jīng)測(cè)得了9只Af

4、和6只Apf數(shù)據(jù)以下：Af: (1.24,1.27)，(1.36,1.74)， (1.38,1.64)，(1.38,1.82)，(1.38,1.90)，(1.40,1.70)，(1.48,1.82)，(1.54,1.82)，(1.56,2.08)；Apf: (1.14,1.82)，(1.18,1.96)，(1.20,1.86)，(1.26,2.00)，(1.28,2.00)，(1.30,1.96).（i）依據(jù)如上資料，怎樣制訂一個(gè)方法，正確地域分兩類蠓蟲(chóng)；（ii）對(duì)觸角和翼長(zhǎng)分別為(1.24,1.80)，(1.28,1.84)與(1.40,2.04)三個(gè)標(biāo)本，用所得到方法加以識(shí)別；（iii）設(shè)

5、Af是寶貴傳粉益蟲(chóng)，Apf是某疾病載體，是否應(yīng)該修改分類方法.神經(jīng)網(wǎng)絡(luò)算法第41頁(yè)2、人口預(yù)測(cè)下表是從北京統(tǒng)計(jì)年鑒中給出 1980-年北京城近郊區(qū)戶籍人口統(tǒng)計(jì)結(jié)果作為樣本數(shù)據(jù)，(1)建立人工神經(jīng)網(wǎng)絡(luò)模型;(2)預(yù)測(cè)年北京城近郊區(qū)戶籍人口.神經(jīng)網(wǎng)絡(luò)算法第42頁(yè)數(shù)據(jù)處理后樣本數(shù)據(jù)：樣本用途樣本組數(shù)輸入一輸入二輸入三輸入四輸出學(xué)習(xí)樣本10.49840.51020.52130.5340.540720.51020.52130.5340.54070.542830.52130.5340.54070.54280.55340.5340.54070.54280.5530.563250.54070.54280.55

6、30.56320.573960.54280.5530.56320.57390.582170.5530.56320.57390.58210.59280.56320.57390.58210.5920.598790.57390.58210.5920.59870.6043100.58210.5920.59870.60430.6095110.5920.59870.60430.60950.6161120.59870.60430.60950.61610.6251130.60430.60950.61610.62510.6318140.60950.61610.62510.63180.6387150.61610.

7、62510.63180.63870.6462160.62510.63180.63870.64620.6518170.63180.63870.64620.65180.6589180.63870.64620.65180.65890.6674190.64620.65180.65890.66740.6786200.65180.65890.66740.67860.6892210.65890.66740.67860.68920.6988220.66740.67860.68920.69880.7072檢驗(yàn)樣本230.6786 0.68920.69880.70720.7132240.6892 0.69880.70720.71320.71