在深度学习中，人工神经网络是一种全面的数学思维的解释方法，分层神经网络的存在促进了图像处理、语音识别、自然语言处理以及生物信息学等领域的发展。然而，分层神经网络的内部分层中包含大量的非线性复杂参数，从中发现或推理知识仍然具有挑战。本文提出了一种从多层神经网络中提取全局简化结构的新方法，使用Python构建一个简单的2-Layered神经网络模型，并对其进行深入解释。本文使用网络分析，对具有相似连接方式的手写字母进行检测，其过程分为三个部分：

(1)网络分解：将训练网络分解为多个独立的小网络，实现任务分解，从而降低时间消耗。

(2)训练评估：使用一个超参数或随机初始参数作为模块化指数，对训练结果的实用性进行分析。

(3)数据分析：对输入层、隐藏层、输出层的结构进行展示，从训练网络中推断信息。

本文首先对提出的算法和其中的数学内涵进行了详细介绍。其次，深入研究混合python代码和数学方程的神经网络编码方式。通过创建带有隐藏层的2-layered神经网络，尽可能的减少时间，从而获得更准确、更高效的输出，此种方式对于两层以上的神经网络也适用。最后，研究了如何扩展所提出的模型，使其更适用于解决复杂的现实问题。

本文的主要贡献是：创建了具有一个或多个隐藏层的2-layered模型，使其能够识别手写图像，并能够扩展到更多领域的任务中。对算法的可行性进行了验证，结果表明。本文提出的网络和模型与其他网络相比，能够更有效的解决多个现实生活中的复杂问题，且能够节约时间。

In Deep Learning Artificial Neural Network could be a comprehensive and mathematical means of explaining a simple neural network of 2-layers through writing one from scratch in Python and at the rear of the scenes of such known breakthroughs. Image process, speech recognition, speech process, and bioinformatics have all seen huge enhancements because of bedded neural networks. However, as a result of a bedded neural network's interior illustration containing many nonlinear and complicated parameters contained in class-conscious layers, it's notwithstanding troublesome to find or interpret understanding from its reasoning. As a result, it is vital to advance a replacement means for comprehending multilayer neural networks. we tend to current a unique technique for extracting a worldwide and simplified structure from a multilayer neural community during this analysis. The cautioned technique detects written letters with comparable affiliation patterns victimization community analysis. we tend to demonstrate its use by inserting it to figure in 3 situations. (1) Network decomposition: It will divide a trained neural network into many tiny freelance networks, reducing computing time and dividing the task. (2) Coaching evaluation: a modularity index will be accustomed to analyzing the suitableness of a trained outcome by the usage of an explicit hyper parameter or randomly generated beginning parameters. (3) Information analysis: it shows the community structure within the input, hidden, and output layers in real-world information, which can be accustomed, to deduce information from a talented neural network.

Firstly, we tend to provide a principle read of the study within the back of the explanation of those breakthroughs.

Secondly, we dived into the writing of neural networks with python code and mathematical equations. During analysis, a two-layered neural network is created with a hidden layer to decrease time and energy succeeding in larger correct and surroundings friendly output, however, the thinking stays constant for additional than 2-layered neural networks.

Finally, we tend to work on however we tend to might publicize our model and build it additionally adjustable for finding advanced real-life issues and our experimental consequences confirmed that our 2-layered network model has to be compelled to attain larger correct and reliable measures as in distinction to totally different community techniques for finding such sophisticated issues.

The 2-layered Neural Network mannequin having one or quite one hidden layer is being created and may be accustomed acknowledges written pictures. The written icon recognition makes use of some set of tasks and implements a range of models and breakthroughs for recognizing written pictures with the usage of the 2-layered Artificial Neural Networks ANNs technique. The practicableness of the improved rule is verified. Such networks and patterns will be accustomed to facilitating and remedying multiple real existing sophisticated issues larger effectively and accurately for this reason saving slot.