文化

Jan 8, 2018 Mar 30, 2018

速记

龙之九子

龙有九个儿子，是跟谁生的？为什么「龙生九子，各不成龙」？ - 豆子的回答 - 知乎
龙有九个儿子，是跟谁生的？为什么「龙生九子，各不成龙」？ - 知乎

一说

龍生九子不成龍，各有所好：
囚牛，龍種，平生好音樂，今胡琴頭上刻獸是其遺像；
睚眦，平生好殺，今刀柄上龍吞口是其遺像；
嘲風，平生好險，今殿角走獸是其遺像；
蒲牢，平生好鳴，今鐘上獸鈕是其遺像；
狻猊，平生好坐，今佛座獅子是其遺像；
霸上，平生好負重，今碑座獸是其遺像；
狴犴，平生好訟，今獄門上獅子頭是其遺像；
贔屭，平生好文，今碑兩旁龍是其遺像；
蚩吻，平生好吞，今殿脊獸頭是其遺像。

李东阳（1447年－1516年）《怀麓堂集》

行于五台

Jan 2, 2018 Mar 30, 2018

出发

旧岁之末，新年伊始，我们来到了五台山。

这是我第一次进行多天的徒步。

去之前的心情还是比较放松的，因为从照片上看，五台山的山势平缓。

路，看起来相当的很好走。

仅有的一丝丝忐忑不安，仅仅是有点低的温度。

CS231n

Dec 24, 2017 Mar 4, 2018

Course notes of CS231n, Convolutional Neural Networks for Visual Recognition.

2. Image Classification pipeline

Challenges
- Viewpoint variation 多视角
- Illumination 光线
- Deformation 变形
- Occlusion 遮挡
- Background Clutter 杂乱的背景
- Intraclass variation 内类多样性
Nearest Neighbor
- Compute the distances of testing images and training images
- \(L_n\) norm
  - \((\sum_{i}{|X^i|^n})^{\frac{1}{n}}\)
- \(L_n\) distance
  - \((\sum_{i}{|X_1^i - X_2^i|^n})^{\frac{1}{n}}\)
- method
  - For each testing image, find the nearest training image
  - Use the label of finding training image as the prediction
- complexity
  - Training: \(O(1)\)
  - Testing: \(O(n)\)

MCB80x

Dec 23, 2017 Mar 4, 2018

MCB80x, the fundamental of neuroscience, is a free, open, online Neuroscience course from HarvardX led by Professor David Cox. Animation by Daniela Sherer + Music by Dan Deacon. Edited by Alex Auriema.

Welcome

Harvard
brain
Three parts
- explore science in Harvard and Boston
- guided interactivity
- DIY science
Modules
- THE ELECTRICAL PROPERTIES OF THE NEURON
  - signal generating, transfering
- NEURONS AND NETWORKS
  - groups of neurons
- THE BRAIN
  - society, intelligence

Matlab

Dec 23, 2017 Apr 15, 2018

matlab

Quick notes

! Matlab start from 1, not 0 !

switch

dataset = 'minist';
switch(dataset):
    case 'cifar-10'
        code;
    case 'minist'
        code;
    otherwise
        error('Unrecognized dataset!')
end

Shell

Dec 20, 2017 May 22, 2018

shell

Links

Quick notes

System information

neofetch, system info
du -sh <file>/<directory>, check the size of file or directory
df -h, display free disk space
- -h, 1k = 1024 byte
- -H, 1k = 1000 byte
uname -s, print kernel name
- Linux
- Darwin
uname -r, linux release number
last -n 50, check last logged in users
lastb -n 50,
pstree -p, process tree
w, show who is logged on and what they are doing

Vim

Dec 18, 2017 Mar 30, 2018

vim

This notes contains useful tips for using Vim or editing and I use spf13-vim.

spf13-vim is a distribution of vim plugins and resources for Vim.
Install spf13-vim: curl http://j.mp/spf13-vim3 -L -o - | sh

Plugins

Surround

Managing all the ‘”[{}]“’ etc.

Characters meaning
- operations
  - d delete
  - c change
  - y add?
- s represents surrounding
- S for visual mode surroundings
- text object (vim editing)
  - w word
  - iw in word
  - aw all word, including space(s)
- surroundings
  - ( left bracket will give a space between it and content
  - ) right won’t
  - b = )
Examples
- yss) surround a line of content with ()
- ysiw<div> surround a word like this <div>nice</div>
- VS<div> surround a line with <div></div>

Forward Learning

Dec 18, 2017 Apr 15, 2018

Papers read during research forward deep learning algorithm.

Quick notes

正推，逆推
- 底层精度高=> 高层精度高
- 相同结构网络，精度高的网络，从中间截取进行分类是否精度也高？
不同粒度特征提取的结合
- 粒度，卷积层数？层数不同的实质是什么不同？
在调整了一些Hyper-paramenter后，大致上可以发现影响更大的参数，比如卷积核数
- 调整优先级：欠拟合 > 过拟合
用CIFAR-10训练时，测试集上的loss会在某次迭代中突然丢失，然后又恢复，形成一个尖刺？
将问题分割成子问题，但试图用深度学习解决的问题，都不太好分割成子问题

Orthogonal Bipolar Target Vectors¹

Can OBV construct a middle target for CNN?

A kind of target representation.

conventional
- BNV - binary: \((0, 0, 1, 0, 0)\)
- BPV - bipolar?: \((-1, -1, 1, -1, -1)\)
OBV - orthogonal bipolar vectors
NOV - Non-Orthogonal Vecotrs
- For fail comparision
- \(V_i=(\overbrace{-1 , \cdots , -1}^{i-1}, 1, \overbrace{-1 , \cdots , -1}^{n-i})\)
- \(cos \theta = \frac{n-2}{n}\)
degraded characters?
- They use degraded license plate images as expirement data. (车牌号)

How to generate OBV from conventional target?

Hinton Course

Dec 18, 2017 Feb 22, 2018

deep learning

Learning notes of Hinton’s Neural Networks for Machine Learning in Coursera.

1. Introduction

1a. Why do we need machine learning

What, Why

We don’t know how to program to solve some problems or the program might be very complicated.
Rules may need to change frequently, like recognizing fraud.
Cheap computation

How

Collect lots of cases with inputs and correct outputs.
ML algorithms takes examples and produces a program to do the job.

Good at

Recognizing patterns
- Objects
- Face
- Spoken words
Recognizing anomalies (unusual)
- Transactions
- Sensor readings in a nuclear power plant
Prediction
- Stock prices, exchange rates
- Movie recommendations

Git

Dec 11, 2017 Jun 30, 2018

git

Some git notes.

Links

Basic configuration

Git config

# List configuration
git config -l
git config --global -l

# Config user's info
git config --global user.name "Mr.Blue"
git config --global user.email "silverhugh.77@gmail.com"

# Delete config
git config --global --unset user.username

# Config editor for git, use vim, nano or so on
git config --global core.editor "vim"
# True way to Commit
#   `git commit` will open an vim editor (set by command before)
#   write the commit message in it and `:x` to commit

文化

速记

龙之九子

行于五台

出发

CS231n

2. Image Classification pipeline

MCB80x

Welcome

Matlab

Quick notes

switch

Shell

Links

Quick notes

System information

Vim

Plugins

Surround

Forward Learning

Quick notes

Orthogonal Bipolar Target Vectors1

Hinton Course

1. Introduction

1a. Why do we need machine learning

Git

Links

Basic configuration

Git config

Orthogonal Bipolar Target Vectors¹