[Tips] ABCDEFGHIJKLMNOPQRSTUVWXYZ

A-ALPHA
B-BETA
C-CHARLIE
D-DELTA
E-ECHO
F-FOX
G-GAMMA
H-HAMMER
I-ICON
J-JAVI
K-KILO
L-LAMBDA
M-MACE
N-NANO
O-OMEGA
P-POD
Q-QUAD
R-ROGUE
S-SIGMA
T-TANGO
U-URBAN
V-VIBRO
W-WAX
X-XRAY
Y-YODA
Z-ZULU

[Work] Solution for WinDVD "Create Overlay Failed"

Just right click desktop, Properties, Settings, Advanced,
Troubleshoot, and slide the hardware acceleration
handle to Full again,

[Work] clustering

Cluster: An Unsupervised Algorithm for Modeling Gaussian Mixtures
Charles A. Bouman

Klustakwik
KlustaKwik is a program for unsupervised classification of multidimensional
continuous data. It arose from a specific need - automatic sorting of neuronal
action potential waveforms (see KD Harris et al, Journal of Neurophysiology
84:401-414,2000), but works for any type of data. We needed a program that
would:
1) Fit a mixture of Gaussians with unconstrained covariance matrices
2) Automatically choose the number of mixture components
3) Be robust against noise
4) Reduce the problem of local minima
5) Run fast on large data sets (up to 100000 points, 48 dimensions)

AutoClass C
AutoClass is an unsupervised Bayesian classification system that seeks a maximum posterior probability classification.

Research in Data Clustering
Welcome to the data clustering page at Michigan State University!

open source clustering software

Efficient Algorithms for K-Means Clustering
Tapas Kanungo, David M. Mount, Nathan S. Netanyahu, Christine D. Piatko, Ruth Silverman, and Angela Y. Wu

ANN: A Library for Approximate Nearest Neighbor Searching
David M. Mount and Sunil Arya

A Fast Implementation of the ISODATA Clustering Algorithm
Nargess Memarsadeghi, David M. Mount, Nathan S. Netanyahu, and Jacqueline Le Moigne

[Work] I enjoy shell programming.

I start to perceive the benefits and essence of shell programming, or more generally the power of Linux/Unix.

These things changed my work and its style. They made some problems tractable, which were before very hard or even impossible to solve.

[Life] 绝望中绽放希望

早上的实验让我几乎彻底失望。
即使采用最费时的方法，我也无法取得好的结果。
我痛苦的看着屏幕上的一个个实验结果发呆。
在思考吗？
脑袋几乎是空白的。
我只听到自己的呼吸声，周围的噪音完全消失。

突然之间，
一道闪光划过脑空，
我发现了一个最重要的基本事实，多么简单而直接的约束！

顿时，我有了强大的信心。
我的心开始歌唱。

面临的下一个问题，是model fitting
这只是技术问题，我会用手头的Levenberg-Marquardt 和 Opt++试试。

接着改了老板编的Highlight detection程序。变成对我完全透明的代码。
我发现这个程序写的是多么的糟糕。一个for循环可以解决的问题，他拷4遍代码。
一些不用的多余变量。 一个没必要的循环检测，很耗费时间。
程序从260多行变成了80多行。简洁，高效。

但还是有一个重要的问题需要回答？
为什么去年的算法还不够？
如果去年的能很好解决现在的问题（会更快， 更少数据）
那为什么要有现在的几乎要累死人的方法。

22：30左右，我终于较为清晰地总结了这些方法的advantages and disadvantages
我也有了清晰的坚实的信心。

I will get it in!

用Sennheiser耳机， 听歌， 放松。


ToDO:
1. 实现想法
2. 难道必须这么麻烦吗？

[Work] trial and error

In this morning I started to test the ideas brewed last night.
All the experiments went failed.
I was totally frustrated.
What is the way out? I cried disperately.

And suddenly I made a basic observation ...

[Work] vision talk

Today Thomas Vetter gave a talk about recent development of morphable model.
The talk is kind of boring. (For general audience, I will say it is quite good.)
It does, however, convey some essence of doing computer vision research.
All most all the computer vision problems look like pretty easy at first glance. But after your serious thinking and investigation, you will find most of the "naive" problems are quite difficult.
A typical approach to these problems is first to define and clarify the problem. Second, we need to build an efficient mathematical model for the problem. Third, we try to solve the problem. Finally, we should test the method with both synthetic and real data.

[Life] Sleepless night

Yesterday I was sad.
Last night was sleepless.
Lots of ideas flooded my mind.
I conveived the future life.

I want to start a company capturing reality.
I want to be a university faculty at the same time.
I will make some breakthroughs for my current project.
...

[Life] A way out?

Life is becoming harder.
I encounter a very difficult problem. Nobody can tell me whether there is a solution.
I must try it by myself and make a way out. Yesterday was one of the darkest days. Today is even worse.

The problem is so nice and attractive. It seems I'm the right person to solve this problem. But how?

The clock steadily ticks and I get nervous sometimes.

[Work] make a batch file

#!/bin/bash

name=$1
nstep=$2

step=0

for pitch in 1024 512 256 128 64 32 16 8; do
#  echo pitch $pitch
  i=0
  while [ $i -lt $nstep ]; do
    echo procam time.txt data\\$name$nstep $pitch $nstep $i
    i=$(( $i+1 ))
  done
  echo " "
done