【正文】 e publication of the book Perceptron, and points out that the sensor can39。它是在現(xiàn)代神經(jīng)科學(xué)研究成果的基礎(chǔ)上提出的，試圖通過模擬大腦神經(jīng)網(wǎng)絡(luò)處理、記憶信息 的方式進(jìn)行信息處理。聯(lián)想 記憶 是非局限性的典型例子。網(wǎng)絡(luò)中處理單元的類型分為三類：輸入單元、輸出單元和隱單元。 1949年，心理學(xué)家提出了突觸聯(lián)系強(qiáng)度可變的設(shè)想。 1986 年進(jìn)行認(rèn)知微觀結(jié)構(gòu)地研究，提出了并行分布處理的理論。反傳網(wǎng)絡(luò)是一種典型的前向網(wǎng)絡(luò)。 Hebb 規(guī)則認(rèn)為學(xué)習(xí)過程最終發(fā)生在神經(jīng)元之間的突觸部位，突觸的聯(lián)系強(qiáng)度隨著突 觸前后神經(jīng)元的活動(dòng)而變化。此時(shí)，學(xué)習(xí)規(guī)律的變化服從連接權(quán)值的演變方程?！按_定性”是因?yàn)樗蓛?nèi)在的原因而不是外來的噪聲或干擾所產(chǎn)生，而“隨機(jī)性”是指其不規(guī)則的、不能預(yù)測(cè)的行為，只可能用統(tǒng)計(jì)的方法描述。自學(xué)習(xí)功能對(duì)于預(yù)測(cè)有特別重要的意義。 利用神經(jīng)基礎(chǔ)理論的研究成果，用數(shù)理方法探索功能更加完善、性能更加優(yōu)越的神經(jīng)網(wǎng)絡(luò)模型，深入研究網(wǎng)絡(luò)算法和性能，如： 穩(wěn)定性 、收斂性、容錯(cuò)性、 魯棒性 等；開發(fā)新的網(wǎng)絡(luò)數(shù)理理論，如：神經(jīng)網(wǎng)絡(luò)動(dòng)力學(xué)、非線性神經(jīng)場(chǎng) 等。將信息幾何應(yīng)用于人工神經(jīng)網(wǎng)絡(luò)的研究，為人工神經(jīng)網(wǎng)絡(luò)的理論研究開辟了新的途徑。目前這方面工作有神經(jīng)網(wǎng)絡(luò)與模糊邏輯、專家系統(tǒng)、遺傳算法、小波分析、混沌、粗集理論、分形理論、證據(jù)理論和灰色系統(tǒng)等的融合。人工神經(jīng)網(wǎng)絡(luò)與其它傳統(tǒng)方法相結(jié)合，將推動(dòng)人工智能和信息處理技術(shù)不斷發(fā)展。 研究方向 神經(jīng)網(wǎng)絡(luò)的研究可以分為理論研究和應(yīng)用研究兩大方面。 優(yōu)越性 人工神經(jīng)網(wǎng)絡(luò)的特點(diǎn)和優(yōu)越性，主要表現(xiàn)在三個(gè)方面： 第一，具有自學(xué)習(xí)功能。 混沌 是一個(gè)相當(dāng)難以精確定義的數(shù)學(xué)概念。使用監(jiān)督學(xué)習(xí)的神經(jīng)網(wǎng)絡(luò)模型有反傳網(wǎng)絡(luò)、感知器等。根據(jù)環(huán)境的變化，對(duì)權(quán)值進(jìn)行調(diào)整，改善系統(tǒng)的行為。這種網(wǎng)絡(luò)實(shí)現(xiàn)信號(hào)從輸入空間到輸出空間的變換，它的信息處理能力來自于簡(jiǎn)單非線性函數(shù)的多次復(fù)合。 1982 年，美國加州工學(xué)院 物理學(xué)家 提出了 Hopfield 神經(jīng) 網(wǎng)格 模型，引入了“計(jì)算能量”概念，給出了網(wǎng)絡(luò)穩(wěn)定性判斷。 發(fā)展歷史 1943 年， 心理學(xué) 家 和數(shù)理 邏輯學(xué) 家 建立了神經(jīng)網(wǎng)絡(luò)和 數(shù)學(xué)模型 ，稱為 MP 模型。非凸性是指這種函數(shù)有多個(gè)極值，故系統(tǒng)具有多個(gè)較穩(wěn)定的平衡態(tài)，這將導(dǎo)致系統(tǒng)演化的多樣性。一個(gè)系統(tǒng)的整體行為不僅取決于單個(gè)神經(jīng)元的特征，而且可能主要由單元之間的相互作用、相互連接所決定。人工神經(jīng)網(wǎng)絡(luò)具有自學(xué)習(xí)和自適應(yīng)的能力，可以通過預(yù)先提供的一批相互對(duì)應(yīng)的輸入－輸出數(shù)據(jù)，分析掌握兩者之間潛在的規(guī)律，最終根據(jù)這些規(guī)律，用新的輸入數(shù)據(jù)來推算輸出結(jié)果，這種學(xué)習(xí)分析的過程被稱為“訓(xùn)練”。英文文獻(xiàn) 英文資料： Artificial neural works (ANNs) to ArtificialNeuralNetworks, abbreviations also referred to as the neural work (NNs) or called connection model (ConnectionistModel), it is a kind of model animals neural work behavior characteristic, distributed parallel information processing algorithm mathematical model. This work rely on the plexity of the system, through the adjustment of mutual connection between nodes internal relations, so as to achieve the purpose of processing information. Artificial neural work has since learning and adaptive ability, can provide in advance of a batch of through mutual correspond of the input/output data, analyze master the law of potential between, according to the final rule, with a new input data to calculate, this study analyzed the output of the process is called the training. Artificial neural work is made of a number of nonlinear interconnected processing unit, adaptive information processing system. It is in the modern neuroscience research results is proposed on the basis of, trying to simulate brain neural work processing, memory information way information processing. Artificial neural work has four basic characteristics: (1) the nonlinear relationship is the nature of the nonlinear mon characteristics. The wisdom of the brain is a kind of nonlinear phenomena. Artificial neurons in the activation or inhibit the two different state, this kind of behavior in mathematics performance for a nonlinear relationship. Has the threshold of neurons in the work formed by the has better properties, can improve the fault tolerance and storage capacity. (2) the limitations a neural work by DuoGe neurons widely usually connected to. A system of the overall behavior depends not only on the characteristics of single neurons, and may mainly by the unit the interaction between the, connected to the. Through a large number of connection between units simulation of the brain limitations. Associative memory is a typical example of limitations. (3) very qualitative artificial neural work is adaptive, selforganizing, learning ability. Neural work not only handling information can have all sorts of change, and in the treatment of the information at the same time, the nonlinear dynamic system itself is changing. Often by iterative process description of the power system evolution. (4) the convexity a system evolution direction, in certain conditions will depend on a particular state function. For example energy function, it is corresponding to the extreme value of the system stable state. The convexity refers to the function extreme value, it has DuoGe DuoGe system has a stable equilibrium state, this will cause the system to the diversity of evolution. Artificial neural work, the unit can mean different neurons process of the object, such as characteristics, letters, concept, or some meaningful abstract model. The type of work processing unit is divided into three categories: input unit, output unit and hidden units. Input unit accept outside the world of signal and data。這種網(wǎng)絡(luò)依靠系統(tǒng)的復(fù)雜程度，通過調(diào)整內(nèi)部大量節(jié)點(diǎn)之間相互連接的關(guān)系，從而達(dá)到處理信息的目的。 （ 2）非局限性 一個(gè) 神經(jīng)網(wǎng)