Note 2 of PRML: Probability Distributions

Density estimation: model the probability distribution $p(x)$ given a finite observation set $x_1,…,x_N$. Parametric density estimation: determine the suitable values of parameters given an observed dataset, such as $(\mu,\sigma^2)$...

Note 1 of PRML: Intro

Supervised learning: classification, regression. Unsupervised learning: clustering, density estimation, visualization. Reinforcement learning: find suitable actions to take by a process of trial and error in a given situation in...

Notes for CS61b2

From lecture 20, CS61b begin to introduce real problems with data structures including set, map, priority queue ADTs. Lecture 20: Disjoint Sets Dynamic connectivity Given a set of...

Notes for CS61b1

Lecture 3: List1, References and Recursion The Golden Rule of Equals (GRoE): Given variables y and x: y = x copies all the bits from x into...

Advice for Applying Machine Learning

This section is mainly about how to debug and optimize a learning algorithm, or, how to architecture or parameters of a learning model. When we build a model making large...

Underfit & Overfit

Underfit & Overfit Before discussing underfit and overfit, we need to define two errors:training error and generalization error. Usually we use training set to train our model and then...

Improving the Way Neural Networks Learn

In this chapter the author explains a suite of techniques which can be used to improve the backpropagation and the way networks learn, including better choice of cost function, named...

Unsupervised Learning

##聚类/clustering Supervised: 给定训练数据x以及对应的标签y进行训练 Unsupervised: 仅给定训练数据x进行训练 K均值算法/K-means 输入 K:聚类个数 训练集:${x^{(1)}, …, x^{(m)}}$ 步骤 Randomly initialize K cluster centroids (mu_1, ..., mu_K) in R^n Repeat{ for i...

Upload a local git repo to GitHub

create a git repository in your GitHub create a local git repository whose name is the same with the git created in 1, and initialized by ‘git init’

Note for Git

集中式版本控制系统: 版本库存放在中央server, 工作时首先联网从中央server取得最新版本, 任务完成后再把改动推送回中央server. 最大的弊端就在于必须联网才能工作. 分布式版本控制系统: 无中央server, 每台电脑均保存完整的版本库, 因此工作时不需要联网, 多人协作时只需要把各自改动推送给对方. 由于不同电脑不在同一LAN, 推送改动不方便, 因此一般也有一台电脑充当”中央server”, 不过其作用仅限于方便交换大家的修改. 安装好git后首先设置name和email: git config --global user.name "name" git config --global...

Mathematics in DeepLearning Book

Linear Algebra Eigen decomposition Eigenvector: a vector satisfying where $A$ is a square matrix, $v$ is called the eigenvector of $A$, $\lambda$...