2022 年 8月随笔档案 - 19376273

摘要：给定a，b，扩展欧几里得算法求得最大公约数的同时，还会给出ax+by=gcd(a,b)的整数解x，y 假设

d_{i - 2} = d_{i - 1} c_{i} + d_{i} d_{i - 1} = d_{i} c_{i + 1} + d_{i + 1}

$d_{i-2}=d_{i-1} c_i+d_i \ d_{i-1}=d_ic_{i+1}+d_{i+1}$ 假设a，b的最大公约数为

g

$g$ ，当某一步的

d_{i - 1} = 0

$d_{i-1}=0$ 时，$1d 阅读全文

posted @ 2022-08-26 15:24 19376273 阅读(28) 评论(0) 推荐(0) 编辑

独立成分分析

摘要：Independent Components Analysis Ambiguity ICA is ambiguous to scaling and permutation. but usually it doesn't matter. As long as the data is not Gauss 阅读全文

posted @ 2022-08-25 22:43 19376273 阅读(33) 评论(0) 推荐(0) 编辑

主成分分析

摘要：Principal Components Analysis Intuition PCA tries to identify the subspace in which the data approximately lies. Intuitively, we choose a direction fo 阅读全文

posted @ 2022-08-25 19:49 19376273 阅读(10) 评论(0) 推荐(0) 编辑

决策树与集成

摘要：Decision Tree Greedy, Top-down, Recurrent Classification Tree misclassification loss is not suitable for decision tree loss, because $$ L(R_p) - (\lam 阅读全文

posted @ 2022-08-24 22:18 19376273 阅读(25) 评论(0) 推荐(0) 编辑

因子分析与EM算法

摘要：Factor Analysis Background when m (number of samples) < n (dimension of samples), the covariance matrix might be singular. So it is necessary to find 阅读全文

posted @ 2022-08-23 18:13 19376273 阅读(30) 评论(0) 推荐(0) 编辑

EM算法

摘要：EM Algorithm Jensen's inequality convex function:

f^{″} (x) \geq 0

$f''(x) \ge 0$ or

H \geq 0

$H \ge 0$ (Hessian matrix when x is a vector)

E [f (x)] \geq f (E X)

$E[f(x)] \ge f(EX)$ EM Algorithm 阅读全文

posted @ 2022-08-22 19:23 19376273 阅读(28) 评论(0) 推荐(0) 编辑

强化学习

摘要：Reinforcement Learning Background Credit Assignment Problem: Explore how actions in an action sequence contribute to the outcome finally. MDP(Markov D 阅读全文

posted @ 2022-08-17 21:08 19376273 阅读(39) 评论(0) 推荐(0) 编辑

学习理论中的特征选取与正则化

摘要：Learning Mathematical Theorem Feature Selection If we have n features, the VC dimension of hypothesis class would be O(n), if n>>m, over-fitting would 阅读全文

posted @ 2022-08-15 16:16 19376273 阅读(25) 评论(0) 推荐(0) 编辑

支持向量机

摘要：Support Vector Machine Two Margins Functional Margin

γ

$\gamma$ and Geometrical Margin

\hat{γ}

$\hat{\gamma}$ $$ \hat{\gamma^{(i)}}=\frac{\gamma^{(i)}}{||w||} $ 阅读全文

posted @ 2022-08-13 10:56 19376273 阅读(34) 评论(0) 推荐(0) 编辑

机器学习中的误差界分析

摘要：Error Theorem in ML Basic Concepts

ε (h) = Generalization Error = E_{(x, y) \sim D} [1 (h (x) \neq y)]

$\varepsilon(h) = \text{Generalization Error} = E_{(x,y)\sim D}[1(h(x) \neq y)]$ $$ \hat{\varepsilon}(h)=\text{ 阅读全文

posted @ 2022-08-12 15:30 19376273 阅读(116) 评论(0) 推荐(0) 编辑

概率密度指数函数族与广义线性模型

摘要：Exponential Family The pdf is as follows,

p (y; η) = b (y) e^{η^{T} T (y) - a (η)}

$p(y;\eta)=b(y) e^{\eta^TT(y)-a(\eta)}$ Mathematical properties: MLE w.r.t.

η

$\eta$ is concave, but negat 阅读全文

posted @ 2022-08-11 16:49 19376273 阅读(97) 评论(0) 推荐(0) 编辑

logistic回归与牛顿法

摘要：Locally Weighted Regression None parametric learning algorithm. Need to keep training data in the memory. Formally fit

θ

$\theta$ to minimize $$ \sum_{i 阅读全文

posted @ 2022-08-10 21:51 19376273 阅读(30) 评论(0) 推荐(0) 编辑

高斯生成模型与朴素贝叶斯

摘要：cs229 Generative Model & Discriminative Model Defination learns P(x|y) : x is feature and y is the class Bayes rule $$ P(y|x)=\frac{P(x|y)P(y)}{P(x)} 阅读全文

posted @ 2022-08-07 23:32 19376273 阅读(40) 评论(0) 推荐(0) 编辑

T5大型语言模型

摘要：T5 and Large Language Model T5 Text-to-Text Transfer Transformer. see all NLP questions as A TEXT-TO-TEXT TASK universal format: task description + se 阅读全文

posted @ 2022-08-05 11:47 19376273 阅读(141) 评论(0) 推荐(0) 编辑

共指消解

摘要：共指消解——Coreference Resolution Coreference Resolution Coreference Resolution: find all the words that refer to the same entity in the world. antecedent 阅读全文

posted @ 2022-08-04 11:46 19376273 阅读(42) 评论(0) 推荐(0) 编辑

向语言模型中添加知识

摘要：Add Knowledge to Language Model Add pre-trained entity embeddings entity linking: map the word occurred in text to the entity in knowledge database. i 阅读全文

posted @ 2022-08-03 20:15 19376273 阅读(62) 评论(0) 推荐(0) 编辑

Stargazer

08 2022 档案

公告

搜索

常用链接

随笔分类

随笔档案

阅读排行榜

评论排行榜

推荐排行榜

最新评论