Skip to content

Commit

Permalink
pf: 调整报告的顺序。
Browse files Browse the repository at this point in the history
  • Loading branch information
heroxbd committed Jul 28, 2024
1 parent 6dc3fe8 commit 159c7a3
Showing 1 changed file with 41 additions and 46 deletions.
87 changes: 41 additions & 46 deletions pd.org
Original file line number Diff line number Diff line change
Expand Up @@ -12273,7 +12273,7 @@ Number of Fisher Scoring iterations: 5
:EXPORT_BEAMER_THEME: Hannover
:EXPORT_BEAMER_FONT_THEME: professionalfonts
:END:
** 复习
** 复习与提示
*** 广义线性回归
1. 关系代数让回归分析变得极其直观,让我们专注于问题的本质
2. 模型的选择极其重要,应当使用客观标准
Expand All @@ -12282,6 +12282,31 @@ Number of Fisher Scoring iterations: 5
3. 广义线性回归,把误差分布从高斯替换为其它指数族分布
+ 把连接函数从恒等替换为非线性函数
+ 几乎可以解决所有日常工作中的非线性问题,惊喜!
*** 数据时代的物理技能
[[./out/Data_Science_VD-migrate.pdf]]
*** 概率是逻辑的扩展 -- Cox 定理
- Laplace: probability theory is nothing but common sense reduced to calculations.
- R. T. Cox, E. T. Jaynes,两位对统计学有重大贡献的物理学家
- Logical interpretation of probability
1. Divisibility and comparability -- The plausibility of a
proposition is a real number and is dependent on information
we have related to the proposition.
2. Common sense -- Plausibilities should vary sensibly with the
assessment of plausibilities in the model.
3. Consistency -- If the plausibility of a proposition can be
derived in many ways, all the results must be equal.
- 概率论的唯一性:
Any system for plausible reasoning that satisfies certain
qualitative requirements intended to ensure consistency with
classical deductive logic and correspondence with commonsense
reasoning is isomorphic to probability theory.

#+latex: {\scriptsize
参考:Van Horn, K.S., 2003. Constructing a logic of plausible
inference: a guide to Cox’s theorem. International Journal of
Approximate Reasoning 34, 3–24.
#+latex: }

** 大作业安排
*** 大作业安排
**** l :BMCOL:
Expand Down Expand Up @@ -12357,7 +12382,7 @@ https://git.tsinghua.edu.cn/physics-data/projects/tpl_junoprobe
- 作业的目标是取得 JUNO 探测器泊松点过程的条件均值函数。

*** likelihood-free inference
- somewhat misleading
- "likelihood-free inference" somewhat misleading
- 从一个没有 likelihood 的模型中,提炼出 likelihood。
** 泊松点过程
*** 随机过程概念
Expand Down Expand Up @@ -12589,11 +12614,12 @@ C_N (s,t)= & \lambda \min(s,t)

** 机器学习
*** 可用 sklearn
- Debian
#+begin_src bash
apt install python3-sklearn-pandas
#+end_src
- Debian
#+begin_src bash
apt install python3-sklearn-pandas
#+end_src

- 下面的算法可以理解为无解析形式的函数逼近。
*** 支持向量机
#+begin_src ein-python :results output :session https://dpcg.g.airelinux.org/user/xubd/lecture-python.ipynb :exports both
### SVM regression
Expand Down Expand Up @@ -12636,6 +12662,11 @@ C_N (s,t)= & \lambda \min(s,t)

自动进行解析微分。

- PyTorch
#+begin_src bash
apt install python3-torch
#+end_src
+ CPU 版本,GPU 版本在 WSL 之内不容易安装,建议申请科协服务器账号。
** 概率编程
*** 概率编程
#+begin_export latex
Expand All @@ -12645,8 +12676,10 @@ C_N (s,t)= & \lambda \min(s,t)
#+end_export
把普通函数关系,自动换成随机变量之间的函数关系。

**** 随机变量
样本空间 \(S=\{e\}\) 的实单值
**** 自动推理机
+ 利用第一阶段大作业成果,直接生成第二阶段大作业的解答程序
+ 语言为 Anglican ,为 Clojure 语言的一种
+ Clojure 语言是 Scheme 语言在 Java 平台上的实现

** 总复习
*** 黑客技能
Expand Down Expand Up @@ -12737,39 +12770,7 @@ C_N (s,t)= & \lambda \min(s,t)
#+end_src

** 下一步学习
*** 数据时代的物理技能
[[./out/Data_Science_VD-migrate.pdf]]
*** 概率是逻辑的扩展 -- Cox 定理
- Laplace: probability theory is nothing but common sense reduced to calculations.
- R. T. Cox, E. T. Jaynes,两位对统计学有重大贡献的物理学家
- Logical interpretation of probability
1. Divisibility and comparability -- The plausibility of a
proposition is a real number and is dependent on information
we have related to the proposition.
2. Common sense -- Plausibilities should vary sensibly with the
assessment of plausibilities in the model.
3. Consistency -- If the plausibility of a proposition can be
derived in many ways, all the results must be equal.
- 概率论的唯一性:
Any system for plausible reasoning that satisfies certain
qualitative requirements intended to ensure consistency with
classical deductive logic and correspondence with commonsense
reasoning is isomorphic to probability theory.

#+latex: {\scriptsize
参考:Van Horn, K.S., 2003. Constructing a logic of plausible
inference: a guide to Cox’s theorem. International Journal of
Approximate Reasoning 34, 3–24.
#+latex: }

*** 以本课程为起点
- 函数式编程:一切都是函数
- 无状态,从而容易从错误中恢复
- MapReduce:分布式大数据系统的开始
- 在之上建立了关系代数:
- Apache Hive,SparkQL,etc.
- 机器学习:与回归分析有同样的程序接口
- 算法上更一般,需要更多的“调参”
**** 竞赛
- probe 大作业 \to Ghost Hunter 2024 中微子数据分析排位赛
- 概率统计分析及量测技术课赛结合
Expand Down Expand Up @@ -12805,12 +12806,6 @@ C_N (s,t)= & \lambda \min(s,t)
+ Structure and Interpretation of Classical Mechanics, MIT 本科课程
+ 分析力学完成后,计划拓展至量子力学和电动力学

#+beamer: \pause
- 自动推理机:
+ 利用第一阶段大作业成果,直接生成第二阶段大作业的解答程序
+ 语言为 Anglican ,为 Clojure 语言的一种
+ Clojure 语言是 Scheme 语言在 Java 平台上的实现

*** 致谢
#+attr_beamer: :overlay <+->
- 顾问 :: 陈晟祺、陈嘉杰
Expand Down

0 comments on commit 159c7a3

Please sign in to comment.