Okay, I need to create a Wikipedia-style document about LeNet under the categories of Technology > Artificial Intelligence > Machine Learning. Let me start by recalling what I know about LeNet.

First, LeNet is one of the earliest convolutional neural networks (CNNs), right? It was developed by Yann LeCun and his team in the 1990s. The main purpose was for handwritten digit recognition, like the MNIST dataset. But wait, when exactly was it created? I think around 1994 or 1998. Maybe I should check that.

The structure of LeNet: It has multiple layers. There's the input layer, then convolutional layers, subsampling (pooling) layers, and fully connected layers. The original version might have been LeNet-5. Let me confirm that. Yes, LeNet-5 is the most well-known version. So I should mention that.

Key components: Convolutional layers use filters to detect features. Subsampling layers reduce spatial dimensions, maybe using average or max pooling. Then fully connected layers for classification. The activation functions—probably sigmoid or tanh? Or was it ReLU? Wait, LeNet used sigmoid and hyperbolic tangent functions. I need to get that right.

Applications: Primarily for OCR (optical character recognition), like reading digits on checks. But how did it perform compared to other models at the time? It was a breakthrough because it showed CNNs could work well for image tasks. Maybe mention its impact on later developments in deep learning.

Challenges and limitations: The original LeNet had limited depth, maybe only 5 layers. Also, training such networks required significant computational resources back then. Data scarcity might have been an issue too. But how did they handle that? They used data augmentation techniques?

Modern relevance: Even though it's old, the principles of LeNet are still taught in CNNs. Maybe mention how modern architectures like VGG or ResNet build upon these ideas.

I need to structure this into sections. The title is "LeNet". Then an overview section introducing what it is, when it was developed, and its significance. Next, a section on architecture details, breaking down each layer type. Then maybe a section on applications and impact. Also, mention the original paper or references.

Wait, the user wants H1 for the title, then H2 sections like Introduction, Architecture, Applications, etc. Maybe also include a "History" section if needed. But I should check the requirements again.

The user specified: 1. Document structure with H1, intro, multiple H2/H3 sections, references. 2. Content quality: accurate, in-depth, clear. 3. Markdown format with proper headings, lists, code blocks if needed. 4. At least 800 characters.

I need to make sure each section is detailed enough. For example, under Architecture, explain each layer type (convolutional, subsampling, fully connected) and their parameters. Maybe include a table of the layers in LeNet-5.

Also, mention the authors: Yann LeCun, Bernhard Boser, et al. The paper was published in 1998? Let me verify that. Yes, "Gradient-Based Learning Applied to Document Recognition" by LeCun et al., 1998.

In terms of technical details, the input size is 32x32 images. Convolutional layers with different filter sizes and numbers. Subsampling using average pooling. Then fully connected layers leading to output.

I should also note that while LeNet was a milestone, it's not used in modern applications due to its simplicity compared to current models. But it's foundational.

For the references section, include the original paper and maybe some key resources or further reading links.

Need to avoid markdown syntax but use proper headings. Wait, the user said "마크다운 형식의 문서" so I need to use H1, H2, etc., with # symbols. Also, check for any technical terms that need explanation, like convolutional layers, pooling, etc.

Make sure the language is clear and professional, avoiding jargon without explanation. For example, when mentioning "convolutional layers," briefly explain what they do in the context of image processing.

Check for accuracy: LeNet-5 has 7 layers (including input), but some sources say 5. Need to clarify that. Also, the number of parameters and how it's structured.

I think I have enough points. Now structure them into sections with appropriate headings, ensure each part is detailed, and meet the length requirement.