① 蛙跳演算法的原理
蛙跳演算法的思想是：在一片濕地中生活著一群青蛙。濕地內離散的分布著許多石頭，青蛙通過尋找不同的石頭進行跳躍去找到食物較多的地方。每隻青蛙個體之間通過文化的交流實現信息的交換。每隻青蛙都具有自己的文化。每隻青蛙的文化被定義為問題的一個解。濕地的整個青蛙群體被分為不同的子群體，每個子群體有著自己的文化，執行局部搜索策略。在子群體中的每個個體有著自己的文化，並且影響著其他個體，也受其他個體的影響，並隨著子群體的進化而進化。當子群體進化到一定階段以後，各個子群體之間再進行思想的交流（全局信息交換）實現子群體間的混合運算，一直到所設置的條件滿足為止。
② 蛙跳演算法的過程
全局搜索過程
步驟l 初始化。確定蛙群的數量、種群以及每個種群的青蛙數。
步驟2 隨機產生初始蛙群，計算各個蛙的適應值。
步驟3 按適應值大小進行降序排序並記錄最好解Px，並且將蛙群分成族群。把F個蛙分配到m個族群Y，Y，Y…,Y中去，每個族群包含n個蛙，從而使得Yk=[X(j),f(j)|X(j)=X(k+m*(j-1), f(j)=f(k+m*(j-1),j=1,…,n,k=1,…,m].這里X（j）表示蛙群中的第j蛙，f(j)表示第j個蛙的目標函數值。
步驟4根據SFLA演算法公式，在每個族群中進行元進化。
步驟5將各個族群進行混合。在每個族群都進行過一輪元進化之後，將各個族群中的蛙重新進行排序和族群劃分並記錄全局最好解Px。
步驟6檢驗計算停止條件。如果滿足了演算法收斂條件，則停止演算法執行過程，否則轉到步驟3。通常而言，如果演算法在連續幾個全局思想交流以後，最好解沒有得到明顯改進則停止演算法。某些情況下，最大函數評價次數也可以作為演算法的停止准則。
局部搜索過程
局部搜索過程是對上述步驟4的進一步展開，具體過程
如下：
步驟4—1設im=O，這里im是族群的計數器。用來與族群總數m進行比較。設iN=0，這里iN是局部進化的計數器，用來與Ls進行比較。
步驟4-2根據式(1)在第l，，1個族群中選擇q個蛙進入子族群，確定Pb和Pw並設im=im+1。
步驟4-3設iN=iN+1。
步驟4—4根據式(2)和式(3)改進子族群中的最差蛙的位置。
步驟4—5如果步驟4—4改進了最差蛙的位置(解)，就用新產生的位置取代最差蛙的位置。否則就採用Px代替式(2)中的PB，重新更新最差蛙的位置。
步驟4—6如果步驟4-5沒有改進最差蛙的位置，則隨機產生一個處於濕地中任何位置的蛙來替代最差的蛙。
不管執行了以上三次跳躍中的任何一次，需重新計算本子群的最優個體Pb和最差個體Pw。
步驟4—7如果iN<LS，則轉到步驟4-3。
步驟4—8如果im<m，則轉到步驟4-2，否則轉到全局搜索過程的步驟5。
演算法停止條件
SFLA通常採用兩種策略來控制演算法的執行時間：
1)在最近的K次全局思想交流過程之後，全局最好解沒有得到明顯的改進；
2)演算法預先定義的函數評價次數已經達到。
3)已有標准測試結果。
無論哪個停止條件得到滿足，演算法都要被強制退出整個循環搜索過程。
③ 什麼是蛙跳演算法
蛙跳演算法（SFLA）是一種全新的後啟發式群體進化演算法，具有高效的計算性能和優良的全局搜索能力。對混合蛙跳演算法的基本原理進行了闡述，針對演算法局部更新策略引起的更新操作前後個體空間位置變化較大，降低收斂速度這一問題，提出了一種基於閾值選擇策略的改進蛙跳演算法。通過不滿足閾值條件的個體分量不予更新的策略，減小了個體空間差異，從而改善了演算法的性能。數值實驗證明了該改進演算法的有效性，並對改進演算法的閾值參數進行了率定
④ 最常見的人工智慧演算法都有哪些
神經網路演算法、蟻群演算法、混合蛙跳演算法、蜂群演算法。
⑤ 仿生智能優化演算法(如何用英語翻譯下）謝謝
Biological modelling intelligence optimization algorithm <dnt> </dnt> each indivial has the experience and the wisdom intelligent body in the biological modelling intelligence optimization algorithm, between the indivial has the interaction mechanism, forms the formidable community wisdom through the interaction to solve the complex problem. The biological modelling intelligence optimization optimization algorithm is one kind of probability searching algorithm essentially, it does not need the question the gradient information, has following is different with the traditional optimization algorithm characteristic:①In the community interaction's indivial is distributional, does not have the direct central main body, not because the indivial indivial will present the breakdown to affect the community to the question solution, will have the strong robustness;②Each indivial can only the sensation partial information, indivial ability or follows the rule to be simple, therefore the community intelligence realizes is simple, is convenient;③The system uses in the expenses which corresponds being few, easy to expand;④From the organization sense, namely the community displays the complex behavior is displays the high intelligence alternately through the simple indivial. Biological modelling intelligence optimization algorithm's these characteristics to overcome difficulties which the optimization design domain faced to provide the powerful support.
<dnt> </dnt>Second, a biological modelling intelligence optimization algorithm common ground analyzes [6] <dnt> the </dnt> several kind of biological modelling intelligence optimization algorithm is simulates the nature living system, total dependence organism own instinct, to optimize its survival condition through unconsciousness optimization behavior to adapt an environment kind of new optimized method, thus has many similar characteristics in the structure and the behavior: 1) is a kind of indefinite algorithm, this kind of uncertainty has manifested the nature physiological mechanism, is follows its randomness to come, when solves certain specific questions must surpass the definite method; 2) is a kind of probability algorithm, its main step includes the random factors, can have more opportunities to gain the globally optimal solution; 3) does not rely on the optimal process the optimized question's strict mathematics nature as well as the objective function and the constraints precise mathematics description; 4) is one kind based on community's intelligent optimization algorithm; 5) has the concealment parallelism, can obtain the great income by the few computations; 6) has appears suddenly the nature, its general objective's completion is in the indivial evolution process appears suddenly in the community; 7) has the evolution, its indivial in complex, stochastic, time-variable environment, through enhances its compatibility unceasingly from the study; 8) has robustness, under the different condition and the environment, manifests the formidable compatibility and the validity. Certainly, because in the nature living system's multiplicity and the complexity, these algorithms also displayed the huge difference. But difference existence, also happen to discuss these biological modelling intelligence optimization algorithm the essential attribute, then obtains the biological modelling intelligence algorithm the unified frame pattern, designed a performance better algorithm to provide the rich material.Second, two biological modelling intelligence optimization algorithm unification frame pattern [7] <dnt> the </dnt> biological modelling intelligence optimization algorithm in aspects and so on structure, research content and method and movement pattern manifested the big similarity, has provided the possibility for the establishment biological modelling intelligence optimization algorithm's unified frame pattern.
<dnt> </dnt> forms the community of the indivial, rests on the specific evolution rule, the iteration proces the renewal community (for example genetic algorithm, ant group algorithm) or the indivial position (for example grain of subgroup algorithm, artificial school of fish algorithm, mix leapfrog algorithm), the optimal solution evolves unceasingly along with the community or the migration appears suddenly, this frame pattern may describe is:
<dnt> </dnt>1) establishes various parameters, proces the initial community and calculates the adaptation value;
<dnt> </dnt>2) acts according to the hypothesis rule, the renewal community or its position, has group of solutions, the computation indivial adaptation value;
<dnt> </dnt>3) obtains the community by the indivial adaptation value comparison the optimal-adaptive value and makes the record;
<dnt> </dnt>4) judges the terminal condition whether to satisfy, if satisfies, conclusion iteration; Otherwise, transfers 2).
<dnt> </dnt> in this frame pattern, the one who decides the algorithm performance is community's renewal rule, these hypothesis rule had decided the indivial behavior standards, have the direct biology foundation, constituted the algorithm to be different with other similar unique essences and the bright characteristic.
<dnt> the </dnt> biological modelling intelligence optimization algorithm sets up together the call-board generally, with records the most superior indivial the historical condition. In algorithm execution each iteration, each indivial comparison own condition and call-board condition, and when own condition is superior with it replacement, causes the call-board to record the historical most superior condition throughout. After algorithm iteration conclusion, may read out the optimal solution from the call-board condition and gain the related information
⑥ 鄒采榮的學術成果
一、發表論文(代表作10篇,部分檢索結果：1997開始至今被收錄SCI 33篇、EI 96篇、CPCI 29篇)：
1.Zou-CR, Plotkin-EI, Swamy-MNS, 2-D Fast Kalman Algorithms for Adaptive Parameter-Estimation of Nonhomogeneous Gaussian Markov Random-Field Model,IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-ANALOG AND DIGITAL SIGNAL PROCESSING , Vol.41, Iss. 10,pp 678-692,1994;
2. Zou-CR, Plotkin-EI, Swamy-MNS, He-ZY. Recursive-in- Order Least-Squares Parameter-Estimation Algorithm for 2-D Noncausal Gaussian Markov Random-Field Model, CIRCUITS SYSTEMS AND SIGNAL PROCESSING Vol.14 Iss.1,pp 87-110,1995;
3.Luo LJ,Lu Y,Zou CR, Image Sequence Macroblock Classification Using Neural Networks, SIGNAL PROCESSING,Vol.69, Iss. 2,pp.191-198,1998;
4.Wang ZH, He ZY, Zou CR, A Generalized Fast Algorithm for N-d Discrete Cosine Transform and Its Application to Motion Picture Coding, IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-ANALOG AND DIGITAL SIGNAL PROCESSING Vol.46, Iss.5, pp.617-627 ,1999;
5. Gao-XQ, Duanmu-CJ, Zou-CR, A Multilevel Successive Elimination Algorithm for Block Matching Motion Estimation, IEEE TRANSACTIONS ON IMAGE PROCESSING,Vol.9, Iss.3, pp.501-504, 2000;
6.Zheng WM, Zhou XY, Zou CR, Facial expression recognition using kernel canonical correlation analysis (KCCA),IEEE TRANSACTIONS ON NEURAL NETWORKS ,Vol.17,pp.233-238,2005;
7.He, Yunhui, Zhao, Li, Zou, Cairong, Face recognition using common faces method,PATTERN RECOGNITION, Vol.39, Iss.11, pp.2218-2222, 2006;
8.Wei Xin, Zhao Li, Zou Cairong,Blind Multiple Access Interference Suppression Algorithm Based on Relaxed Subgradient Projection for DS/CDMA Systems, CIRCUITS SYSTEMS AND SIGNAL PROCESSING Vol.29, Iss.4 pp.769-780,2010;
9.Sun Ning, Ji Zhen-hai, Zou Cai-rong , Two-dimensional Canonical Correlation Analysis and Its Application in Small Sample Size Face Recognition, NEURAL COMPUTING & APPLICATIONS,Vol.19,Iss.3,pp.377-382,2010;
10.Cairong Zou, Chengwei Huang, Dong Han, Li Zhao. Detecting Practical Speech Emotion in a Cognitive Task, Computer Communications and Networks (ICCCN), 2011 Proceedings of 20th International Conference on, Maui, HI, USA, 2011。
二、授權發明專利：
1.一種基於二維偏最小二乘法的面部表情識別方法;
2.一種基於加權主動形狀建模的人臉特徵定位方法;
3.一種人臉身份和表情的同步識別方法;
4.一種雷達脈內調制信號的特徵提取方法;
5.一種基於改進Fukunage-koontz變換的語音情感識別系統;
6.淺海水聲通信系統的間接自適應均衡方法;
7.水聲網路中的節能的媒質訪問控制方法;
8.基於JND和AR模型的感知視頻壓縮方法;
9.基於改進的VLS的立體視頻編碼方法;
10.一種基於支持矢量基的語音情感識別方法;
11.一種適用於中國數字電視地面廣播國家標準的同步方法;
12.基於分數傅里葉變換的二維維納濾波的取證語音增強方法;
13.一種加權次梯度投影的數字助聽器回聲路徑估計方法;
14.基於改進BP演算法的中間視合成方法;
15.一種基於CDMA水聲網路的媒質訪問控制方法。
三、已授權實用新型專利3項：
1.超高精度壓力計量校準儀;
2.新型滴眼裝置;
3.帶有信息檢索的智能電視終端。
四、已授權外觀設計專利1項：
1.活頁式電子樂譜。
五、已登記軟體著作權1項：
1.MusicPro電子樂譜系統軟體V1.0(登記號：2008SR38814)。
六、正在申請並受理發明專利15項：
1.基於樂符知識及雙投影法的樂符基元分割方法;
2.一種基於心電信號與語音信號的雙模態情感識別方法;
3.一種針對煩躁情緒的可據判的自動語音情感識別方法;
4.基於情感對特徵優化的語音情感分類方法;
5.一種分數傅里葉變換上的時頻域掩蔽信息隱藏方法;
6.一種基於分數傅里葉變換域的隱秘信號同步方法;
7.基於分段投影與樂符結構的譜線檢測及刪除方法;
8.基於行遊程鄰接表的樂譜快速連通域分析方法;
9.一種基於多變數統計的助聽器聲源定位方法;
10.一種基於壓縮感測的助聽器聲源定位方法;
11.一種認知無線電功率控制方法;
12.一種基於雲理論與分子動力學模擬的混合蛙跳演算法;
13.基於特徵空間自適應投影的語音情感識別方法;
14.一種跨語言的語音情感識別方法;
15.負面情緒檢測中的基於上下文修正的語音情感識別方法。
七、科技獲獎：
1.「基於面部表情和情感語音的兒童情緒能力分析與分類的研究」獲2009年度江蘇省科學技術進步二等獎.排名 第一;
2.「情感特徵分析與識別的理論與應用」獲2008年教育部自然科學二等獎 排名 第一;
3.「多維數字信號處理的理論與應用研究」 獲1998年國家教育部科技進步(基礎類)二等獎 排名第三;
4.「盲信號模型參數估計的方法研究」獲2000年中國高校科學技術獎勵委員會二等獎 排名第四;
5.「小波與濾波器組的理論及其應用研究」獲2006年教育部自然科學二等獎 排名 第三;
6.「神經網路理論及其智能信息處理應用基礎」獲1998年國家教育部科技進步(基礎類)一等獎 排名第二十二。
八、參編著作：
1.《多維數字信號處理》，何振亞主編，國防工業出版社 1995。獲江蘇省優秀教材一等獎、教育部2001年優秀教材一等獎。
⑦ 仿生演算法有那些
仿生演算法———遺傳演算法、蟻群演算法和混合蛙跳演算法
