---
theme: ./
layout: cover
class: text-white
logoUrl: https://picx.zhimg.com/80/v2-32c67121f7b3b7395aec3ad20fb97713_1440w.png
topic: 《影像与智能科研实践》课程讲义
coverOrganization: Miracle Lab
coverAuthor: [Kevin Zhou, 锦恢]
coverAuthorUrl: [https://baike.baidu.com/item/%E5%91%A8%E5%B0%91%E5%8D%8E/61922413, https://kirigaya.cn/about]
coverOccasion: 明德楼 C307
coverMedia: https://picx.zhimg.com/80/v2-a9d1b704a4302f984012a03a67ffa61c_1440w.png
---
# 科研实践讲演
AI Agent 基础技术讲解
---
layout: table-of-contents
hideInToc: false
---
# 讲演大纲
---
layout: index
indexEntries:
- { title: "image 布局", uri: 4 }
- { title: "text-image 布局", uri: 5 }
- { title: "GitHub", uri: "https://github.com/alexanderdavide/slidev-theme-academic" }
- { title: "npm", uri: "https://www.npmjs.com/package/slidev-theme-academic" }
- { title: "Slidev", uri: "https://sli.dev" }
---
# index 布局
`index` 用作列表，`uri` 可以是 slidev 的页面 id，也可以是外部链接。
```yaml
layout: index
indexEntries:
- { title: "image 布局", uri: 4 }
- { title: "text-image 布局", uri: 5 }
- { title: "GitHub", uri: "https://github.com/alexanderdavide/slidev-theme-academic" }
- { title: "npm", uri: "https://www.npmjs.com/package/slidev-theme-academic" }
- { title: "Slidev", uri: "https://sli.dev" }
```
---
layout: image
caption: Ender Magnolia Development
footnodeNumber: 1
media: https://pica.zhimg.com/80/v2-9df4eb94bfda11be1874fc6af61087ac_1440w.jpeg
---
# image 布局
https://www.endermagnolia.com/en/
---
layout: text-image
caption: Curated cover image for Slidev
footnodeNumber: 1
media: https://pic1.zhimg.com/80/v2-ca2a555a68993804ed948795aa5ea4f2_1440w.webp
---
# text-image 布局
- 确保图表默认展示美观
- 支持将图表放置在左侧或右侧
- 提供可选的图表标题功能
## frontmatter
```yaml
layout: text-image
caption: Curated cover image for Slidev
footnodeNumber: 1
media: https://pic1.zhimg.com/80/v2-ca2a555a68993804ed948795aa5ea4f2_1440w.webp
```
---
# 默认布局
今天，我们学习如何使用 numpy 进行数组运算
```python
import numpy as np
# 创建数组
a = np.array([1, 2, 3, 4])
b = np.array([10, 20, 30, 40])
# 基本运算
print("a + b =", a + b) # 向量加法
print("a - b =", a - b) # 向量减法
print("a * b =", a * b) # 元素逐个相乘
print("b / a =", b / a) # 元素逐个相除
# 广播机制
c = np.array([[1], [2], [3]]) # 3x1 向量
d = np.array([10, 20, 30]) # 1x3 向量
print("Broadcasting:\n", c + d)
# 矩阵运算
A = np.array([[1, 2], [3, 4]])
B = np.array([[2, 0], [1, 2]])
print("矩阵乘法:\n", A @ B)
# 统计运算
print("a 的和 =", a.sum())
print("b 的平均值 =", b.mean())
print("A 的最大值 =", A.max())
```
---
# MCP 服务器
```mermaid
flowchart LR
subgraph Backend [Agent 后端服务]
A[POST 请求]
MCP_Client[MCP Client
内置Agent循环]
A -- 根据参数 --> MCP_Client
end
%% MCP Servers
subgraph MCP_Servers [MCP Server 层]
Prompt_Server[Prompt]
Tool_Server[Tool]
Resource_Server[Resource]
end
%% MCP协议通信
MCP_Client <-.->|MCP协议| Prompt_Server
MCP_Client <-.->|MCP协议| Tool_Server
MCP_Client <-.->|MCP协议| Resource_Server
%% 数据存储
DB[(数据库)] -- 通过Resource Server访问 --> Resource_Server
Tools[外部工具/API] -- 通过Tool Server调用 --> Tool_Server
```
---
# Raden 变换
二维逆傅里叶变换给出 $f(x,y)$：
$f(x,y) = \iint_{-\infty}^{\infty} F(k_x,k_y)\, e^{j2\pi (k_x x + k_y y)}\, dk_x\,dk_y.$
换成极坐标 $(k_x,k_y) = (\omega\cos\theta, \omega\sin\theta)$，有 $dk_x\,dk_y = |\omega|\,d\omega\,d\theta$。于是
$\begin{aligned} f(x,y) &= \int_{0}^{\pi}\int_{-\infty}^{\infty} F(\omega\cos\theta,\omega\sin\theta)\, e^{j2\pi \omega (x\cos\theta + y\sin\theta)}\,|\omega|\,d\omega\,d\theta \\ &= \int_{0}^{\pi}\int_{-\infty}^{\infty} P_\theta(\omega)\, |\omega|\, e^{j2\pi \omega (x\cos\theta + y\sin\theta)}\, d\omega\, d\theta, \end{aligned}$
其中我们用到了投影-切片定理将 $F$ 换成 $P_\theta$。
---
layout: center
class: "text-center"
---
# Footnotes & Footnote
Smart person
---
layout: center
class: "text-center"
---
# Pagination
---
layout: end
---
# Thank you
zhelonghuang@qq.com
https://kirigaya.cn/about