In "The Innovator's Dilemma", the arthur thinks when a company encounters a disruptive technology,
在創新的兩難中,作者認為一家公司在面對突破性科技時,最重要的不是改善突破性科技的效果,把突破性科技應用在現有的市場中。而是替突破性科技找到新的市場。
在Pattern Recognition的研究領域裡,研究人員一直在想辦法提升辨識的準確率,使得PR可以有效的被應用在人們的日常生活中。但是「創新的兩難」替我們指引出了另一個新的方向。 也許最重要的不是提升準確率,而是找到一個新的應用,而這個應用就算是準確率不高也沒有關係,因為有自動化的辨識,所以提供了和現有的技術不同的性質。
數位相機的人臉辨識就是一個例子。如果人臉有抓到,很好。但是如果沒抓到呢?也沒關係,反正相片還是拍的出來。
Sep 16, 2009
Application matters!
應用才重要
Posted by Lono at 21:43 0 comments
Labels: Research
Sep 15, 2009
How to minimize the number of logging data
In the last week, I proposed an idea to my boss about minimizing logging data.
The idea is simple. Assume that we have a deterministic system. We want to debug the system, so we need some logging data to analyze the problem. Since the system is deterministic, if we give it the same input, it will respond the same output. So we only need to log the data coming into the system. If we made the system deterministic, we could avoid logging in the internal of the system. It can save a lot of logging bandwidth.
However, my boss thought it is impossible. He said, "It's impossible to log all incoming data of the system, because we cannot log all incoming electromagnetic waves. We can only log the data like the XOR of the electromagnetic waves." I tried to explain to him that we only care about digital systems, therefore we can focus on 0's and 1's, but he still very persisted that electromagnetic waves are impossible to log. It's near the lunch time, so I quickly ended this unpleasant conversation and went for my lunch.
After few days, I started to wonder "what is the XOR of electromagnetic waves?". XOR is a Boolean operation, how to apply it to continuous signals? Another thing I wondered is "what is the maximum input of a 3.5G mobile phone system ?" I thought the answer should be below 15Mb/s, because the maximum data rate of HSDPA is 7.2Mb/s. The HSDPA protocol overhead should not exceed the amount of data. If our system were deterministic, the minimum number of logging data rate should be below 15Mb/s. As a result, the data rate is ok to transfer by merely an USB cable, am I correct?
Posted by Lono at 23:10 2 comments
Sep 7, 2009
How to get WHQL certificate for your driver
If you need to get WHQL from Microsoft for your driver, you need to pass Windows Logo Test. First you need a PC with Windows Server 2003 installed.
Then follow the steps in DTMHowTo.
When you pass the logo test ( sometimes you need to the support from Microsoft ), see Winqual Faq to know how to submit your result to Microsoft.
Posted by Lono at 07:42 0 comments
Labels: Driver
Sep 3, 2009
Summary of MediaTek Inc.
todo
What my boss said today
1. Our product is good, because we are successful on the market, and our performance compared to competitors are good
2. We don't do unit test and code review unless we feel painful because of them
3. Unit test scale(on a module or multiple module or a function?)
4. Tools can have its innovation, not just support
The people in MediaTek thinks they can work here because they are superior than others. They think their job is the best in the world. One day, a colleague said to me that "If you leave this company, you will never get a better job. At that time, you have to beg to return."
I hardly agree with what he said. Indeed, thanks to the tax policy in Taiwan, MediaTek's compensation is considered the best in Taiwan. However, I don't think good compensation is the ONLY criteria for a good job. Of course, money is important, but it's more important to consider the money which you get and the money which you will get at the same time. I think it's better to judge a people's value by what he is going to be paid in his next job than what he is paid in his current job. When you are fired by your boss and want to find a new job, you cannot just put the number of your previous compensation on your resume and convince your future boss that you are worth that much. You have to prove it by what you have learned and what you have done in your previous jobs. These are more important than just good compensation.
For myself, I found my current job hardly add any eye-catching acheivements to my resume. I need to do something to differentiate myself from fellow engineers. In a competitive job market, the secret to succeed is not only do your job right, but also choose the right job to do.
Many people think that I quit a high compensation job is unrealistic, but most of time, I found myself very reasonable.
to summarize:
1. This article will become like Code Horror style
2. I am glad to leave
Posted by Lono at 17:19 0 comments
Labels: Summary
Aug 27, 2009
Summary of The Innovator's Dilemma by Clayton M. Christensen

Mr. Christensen tries to answer "why the best firms can fail" in this book. After the analysis from HDD industry, he concludes that the best firms can fail when they are unable to deal with disruptive technologies correctly. My quick summary is below:
A disruptive technology has three properties:
1. The performance of the technology is going to exceed the demands of certain market segments
2. The technology distinguish itself with some new features which appeal to a group of new customers.
3. The new market segments are too small to attract the attention of the leading companies of the existing technology.
What will happen when a disruptive technology emerges?
1. The startup who finds the new market for it has the "first mover advantage". Those who find have 30% survival rate.
2. The startup who focuses on improving the performance to meet the existing demands is going to be knock out of the market.
3. Existing companies will still focus on existing customers since they are more profitable.
4. The useful techniques like PDCA or Customer-Oriented Management help existing companies compete with one another, but they also prevent existing companies to enter the new market.
What should a existing company do?
1. A big company is managed by existing customers, process and values, not managers. Establish a new department to manage the disruptive technology if the required values and process are different from the existing one.
2. Do not try to improve the performance of a disruptive technology, but to find a new market for it. The new technology will grow itself with the new customers.
3. It's impossible to predict where the new market is. One has to try several times to find it. Don't put all resources in a single try.
4. When looking for new markets, don't hear what customers say, but see what they actually use the product.
Example:
1. YouTube: It's a example of a disruptive technology. The quality of its video hosting service exceed what people demands. It provides a new type of web service. In the beginning, the profit of such service is too small to attract attention of any existing companies. Nevertheless, it grows big enough to be acquired by Google. A good end.
2. Netbook: It's commercialized first by Asus. The performance of netbook will compete with notebooks in the future. However, Asus does not have any first mover advantages. The followers like Acer and HP got ahead of ASUS very quickly. Why?
Just think about property 2 and 3 above. The people who buy netbook also have notebooks. There are no new customers. The existing notebook manufacturers join the battlefield quickly. Therefore, there are no first mover advantages for ASUS. Netbook is not a disruptive technology but a sustaining one.
3. SSD: Compared with HDD, the capacity per dollar of SDD is very low. However, the value increases every year. When it hits the sweet point, HDD will be wiped out from personal computer market. The first mover advantage is very clear. No leading HDD manufactures have influence on SDD markets.
Posted by Lono at 09:26 0 comments
Labels: Summary
Aug 24, 2009
Clothes
1. 圓領條紋T單穿不會很單薄
2. 黑T可搭黑白格子衫
3. 黃T可搭黃衫
4. T + 衫的重點是協調,而非對比
todo:拍照
Posted by Lono at 09:24 0 comments
Labels: Life
Jun 28, 2009
VIM Color Theme (Ayende)

Oren Eini made a great color theme for Visual Studio. I ported it to VIM. You can get it here
Here is a list of my previous VIM themes:
1. AF.vim

2.VividChalk.vim (it's specialized for Ruby)

Posted by Lono at 05:11 0 comments
