人工神经网络导论_神经网络导论

心已赠人 2022-12-07 13:30 238阅读 0赞

![ee73f9e74351c8b0e456365d25183d81.png][]

人工神经网络导论

Here you will get an introduction to neural networks in the field of data science.

在这里，您将对数据科学领域的神经网络进行介绍。

Neural networks are similar to biological neural network. Biological neural network is collection of biological neurons in human brain similarly Neural network is collection of nodes called Artificial neurons.

神经网络类似于生物神经网络。 生物神经网络类似地是人脑中生物神经元的集合。神经网络是称为人工神经元的节点的集合。

Neural networks are based on non-task specific programming concepts like in image recognition they learn to resemble images by analyzing sample images labeled with name as “car” or “no car” and by using such sample example they identify car in other images. They need not to acquire any knowledge regarding car like it has engine, four wheels, shape and so on. They generate their own relevant characteristics from the process of their learning material.

神经网络基于非任务特定的编程概念，例如在图像识别中，它们通过分析标记为“ car”或“ no car”的样本图像来学习类似于图像的方式，并通过使用此类样本示例将其识别为其他图像中的汽车。 他们不需要获得有关汽车的任何知识，例如它的发动机，四个车轮，形状等。 他们从学习材料的过程中产生自己的相关特征。

Neural Network is a feature of artificial intelligence that efforts to copy the way human brain works. Neural network perform its operation by connecting the processing elements rather than doing all computations that manipulate zeros and ones in digital model

神经网络是人工智能的一项功能，旨在复制人脑的工作方式。 神经网络通过连接处理元件来执行其操作，而不是执行在数字模型中操纵零和一的所有计算

Neural network is used to represent relationships between complex input/output and also it is capable to capture data the same way human brain works. This idea for the development of neural network technology arises from the desire to perform all task intelligently similar to human brain and develop an artificial system to perform all this task.

神经网络用于表示复杂输入/输出之间的关系，并且能够以与人脑相同的方式捕获数据。 神经网络技术发展的这种想法源于对智能地执行类似于人脑的所有任务并开发一种人工系统来执行所有任务的需求。

Neural network gains knowledge through learning. Synaptic weight i.e. inter-neuron connection strengths are used to store a neural network’s knowledge.

神经网络通过学习获得知识。 突触权重，即神经元间的连接强度，用于存储神经网络的知识。

### 神经网络的基本组织 **(**Basic Organization of Neural Network**)** ###

In Neural network, each connection called synapse between nodes or artificial neurons is used to transmit signal from one another and the receiving neuron called postsynaptic can process the signal and information and thereafter signal neurons connected to it.

在神经网络中，节点或人工神经元之间的每个称为突触的连接都用于相互传输信号，而称为突触后的接收神经元可以处理信号和信息，然后再与之连接的信号神经元。

Generally, neurons are arranged in layers. Different layers may perform different transformation with their inputs. Signals travelling from input i.e. the first layer to output i.e. the last layer, possibly have to traverse the layers multiple times.

通常，神经元分层排列。 不同的层可能会对它们的输入执行不同的转换。 从输入即第一层传播到输出即最后一层的信号可能必须多次遍历这些层。

In neural networks the sending signal (synapse) is real number and output is calculated by non-linear function. Output of each neuron is calculated by non-linear function of the sum of its inputs. Synapses and artificial neurons have weights that are adjustable as learning proceeds. The weight of neurons increases or decreases the signal strength that it has to send over synapse .

在神经网络中，发送信号(突触)是实数，并且输出是通过非线性函数计算的。 每个神经元的输出是通过其输入总和的非线性函数来计算的。 突触和人工神经元的权重可以随着学习的进行而调整。 神经元的重量增加或减少了它必须通过突触传递的信号强度。

Neural network solve the problem the same way human brain do. Neural network have been used in computer speech recognition, video games, machine translation and medical diagnosis.

神经网络以与人类大脑相同的方式解决问题。 神经网络已用于计算机语音识别，视频游戏，机器翻译和医学诊断中。

![Introduction To Neural Networks][]

[Image Source][]

[图片来源][Image Source]

### 神经网络的类型 **(**Types of Neural Network**)** ###

#### 1.前馈神经网络 **(**1. Feedforward Neural Network**)** ####

Feedforward neural network is simplest of all neural network. It has no cycles or loops in its network that is it moves information in one direction only. The data in feedforward network moves from input to output nodes passing through hidden layer if there is any hidden layer in between the input and output node.

前馈神经网络是所有神经网络中最简单的。 它的网络中没有循环或循环，仅在一个方向上移动信息。 如果输入和输出节点之间存在任何隐藏层，则前馈网络中的数据将从输入节点移动到通过隐藏层的节点。

#### 2.径向基函数神经网络 **(**2. Radial Basis Function Neural Network**)** ####

Radial basis function neural network is highly spontaneous neural network. It is the first choice in multidimensional space during interpolation. In radial basis function neural network, each neuron saves an example from training set as a “prototype”. Radial basis function has advantage that it doesn’t suffer from local minima offered by linearity involved in functioning of it.

径向基函数神经网络是高度自发的神经网络。 它是插值过程中多维空间的首选。 在径向基函数神经网络中，每个神经元将训练集中的一个示例保存为“原型”。 径向基函数的优点是不会受到函数线性涉及的局部最小值的影响。

#### 3. Kohonen自组织神经网络 **(**3. Kohonen Self-Organizing Neural Network**)** ####

Kohonen self-organizing neural network is used to visualize high dimensional data as low dimensional views. It is invented by Teuvo Kohonen. Kohonen self-organizing  neural network is used to describe hidden structures in it by performing functions on unlabeled data. competitive learning is applied by self-organizing neural network to a set of input data not error correction learning applied by other neural network.

Kohonen自组织神经网络用于将高维数据可视化为低维视图。 它是由Teuvo Kohonen发明的。 Kohonen自组织神经网络用于通过对未标记数据执行功能来描述其中的隐藏结构。 竞争性学习是通过自组织神经网络应用于一组输入数据，而不是其他神经网络进行的纠错学习。

#### 4.递归神经网络 **(** 4. Recurrent Neural Network**)** ####

Recurrent Neural Network is used for parallel and sequential computation it is used to compute each and every thing similar to traditional computer. Recurrent neural  network  works  similar to human brain, it is a large feedback network of connected neurons that can translate a input stream into a sequence of motor outputs. A recurrent neural network (RNN) use their internal memory to process sequencing of inputs. In recurrent neural network, connections between units form a directed cycle. Recurrent neural network model each vector from sequence of input stream vectors one at time. This allows  the network to retain its state during modeling of each input vector across the window of input vectors.

递归神经网络用于并行和顺序计算，它用于计算与传统计算机类似的每件事。 递归神经网络的工作原理类似于人脑，它是一个连接神经元的大型反馈网络，可以将输入流转换为一系列运动输出。 递归神经网络(RNN)使用其内部存储器来处理输入的排序。 在递归神经网络中，单元之间的连接形成有向循环。 递归神经网络一次从输入流向量序列中对每个向量建模。 这允许网络在跨输入向量窗口的每个输入向量建模期间保留其状态。

#### 5.模块化神经网络 **(**5. Modular Neural Network**)** ####

Modular neural network consist of series of independent neural networks that are operated by intermediary. Each independent neural network act as a module and works with separate input that network hopes to perform by accomplishing subtask. The intermediary accepts the output of each module and helps them to process the final output for modular neural network. Also the independent neural network i.e. the modules do not interact with each other.

模块化神经网络由一系列由中介操作的独立神经网络组成。 每个独立的神经网络都充当模块，并与网络希望通过完成子任务来执行的单独输入一起工作。 中介接受每个模块的输出，并帮助他们处理模块化神经网络的最终输出。 同样，独立的神经网络，即模块之间不相互影响。

#### 6.物理神经网络 **(**6. Physical Neural Network**)** ####

The physical neural network aims to focus on physical hardware along with the software while simulating the neural network. A resistance material that is electrically adjustable is used for emulating the function of neural synapse. When the physical hardware emulates the neurons, the software emulates the neural network.

物理神经网络旨在在仿真神经网络时将重点放在物理硬件以及软件上。 电可调节的电阻材料用于模拟神经突触的功能。 当物理硬件模拟神经元时，软件将模拟神经网络。

### 神经网络的应用 **(**Applications of Neural Networks**)** ###

*  Speech Recognition
    
    语音识别
 *  Character Recognition
    
    字符识别
 *  Signature Verification Application
    
    签名验证申请
 *  Human Face Recognition
    
    人脸识别

**About Author:**

**关于作者：**

*Shubham Sharma, currently working as Analytics Engineer in Data Science Domain. Has around 2+ years of experience in Data Science. Skilled in Python, Pandas, Anaconda, Tensorflow, Keras, Scikit learn, Numpy, Scipy, Microsoft Excel, SQL, Cassandra and Statistical Data Analysis, Hadoop, Hive, Pig, Spark, Pyspark. Connect with him at [\[email protected\]][email_protected]*

*Shubham Sharma，目前在数据科学领域担任分析工程师。* *在数据科学领域拥有大约2年以上的经验。* *精通Python，Pandas，Anaconda，Tensorflow，Keras，Scikit学习，Numpy，Scipy，Microsoft Excel，SQL，Cassandra和统计数据分析，Hadoop，Hive，Pig，Spark，Pyspark。* *通过[\[电子邮件保护\]][email_protected]与他联系*

Comment below if you found any information incorrect or have doubts related to above tutorial for introduction to neural networks.

如果发现任何信息不正确或对上述神经网络入门教程有疑问，请在下面评论。

> 翻译自: [https://www.thecrazyprogrammer.com/2017/12/introduction-neural-networks.html][https_www.thecrazyprogrammer.com_2017_12_introduction-neural-networks.html]

人工神经网络导论

[ee73f9e74351c8b0e456365d25183d81.png]: /images/20221123/df1ce8fcf8ac4fc593735ef7650b52ae.png
[Introduction To Neural Networks]: /images/20221123/d5050d917318427486799c9cf734ae8a.png
[Image Source]: http://ecee.colorado.edu/~ecen4831/lectures/NNdemo.html
[email_protected]: https://www.thecrazyprogrammer.com/cdn-cgi/l/email-protection
[https_www.thecrazyprogrammer.com_2017_12_introduction-neural-networks.html]: https://www.thecrazyprogrammer.com/2017/12/introduction-neural-networks.html