Mission to Books.

I'm a poor writer. Terribly poor. But, I wish some day I'll write a book for my feelings, wisdom, experience, whatever else. As an engineer, I have written a few papers, which are not so good but gave me another title from Mr. to Dr. A paper is a container of knolwedge, I think. From very raw data, we extract patterns, so-called infomration, then publicly express them via paper publication. The expressed and collected pieces of information can be called as knowledge. Yes, I have been working to make knowledge from data. However, I wish to express my feelings, experiences, and ideas, which may not be knowledge to be published. I named those things as widsom. Why I want to write book(s) is it is a container for widsom. All of the people around me know that I'm not talkative. Usually, I make less than 100 words in a day. That does not mean that I don't like to speak or I have nothing to speak. Actually, I have many things to talk and want to express myself as I am. Why I am attached to this blog -- anonymous web pages -- is partialy due to such a desire. Not to prepare my writing skill nor to collection stories to fill my books if I'll. Just to express myself to everybody. I know. Nobody read these messages. But, I don't care. I don't expect you read this, but I just write my feelings, what I thought. That's all. ... Anyway, one day I am a writer. I'll have my books on my shelter. My widsom will be near me. Yeah, the mission is set. ... I'm a writer.

A Simple Business Idea: A Smart Music Player

Someone might already implement and/or get a patent of this idea, so that I was hesitate to write this. But, if not, you are the one to do this. An uncompressed song requires about 40 MB, but it becomes around 4 to 10 MB according to the compression rate. This is the underlying idea of MP3. The compression means cutting-off very very important information from the original source. This may not mean to a dummy listener like me, but definitely means experts (or protuers). I'm OK with the compressed song requiring smaller space. But, I'm not OK the quality of those compressed songs. Why not interpolate the cut-off part of them by a music player (both hardware and software). This may a problem of semi-supervised learning or not. By a simple regression, we can interpolate the plat part (no peak by cutting-off) with previous information (gradually increasing/decreasing). The interpolation must be applied to both sides -- increasing part and decreasing part, or vice versa. I want to call this 'software interpolation/restoration', because the music is interpolated or restored by software, when it is running. That results in the music file is still compressed but the live listening is no more compressed. I'm not an expert on sound signal.. so I don't know how to extract the sound signal. This means that I have the idea but I will no more implement it. So, you are the one. Make it as your idea. Or, apply it to your business if you have concerns.

Research Ideas: Brain Inspires Me Again.

As a scientific researcher (my recent concern is data mining and machine learning), brain may make a quantum jump in that field. Traditionally, the brain/nerve system inspired the neural networks. As it did, it can do it again in different ways. It is well known that human brain consists of two spheres -- right brain and left brain. Roughly speaking, intelligence things come from the left and emontional things from the right. In addition, the right views a thing in an integrated manner, while the left views the same thing very specifically. The different functionality makes human see both every tree and whole trees. This property is useful to make an intelligent machine learning algorithm for various applications. We may make two different classifiers, for short, one predicts the general trend whereas the other predicts the specific variations. And then, by ensembling them, we may get a more accurate classifier. Take the stock market prediction for example. (I know that it is nearly impossible to predict long-term stock price.) One classifier just catches the long-term trend and the other catches the short-term fluctuation. The ensemble classifier may get a right decision. Another example can be taken from seasonal market share data -- yearly share may gradually increase but the seasonal factor makes the share fluctuated. In this case, one classifier is responsible for the yearly increase while the other is for the seasonal fluctuaion.

In addition, one brain can interference with others. For instance, if a person feels sad, then other persons in the same room also feel sad. This idea can also be applied to improve or design a new machine learning algorithm. Interference machine... sounds good.

Business Ideas: Personal Advertisement System

I'm not sure that the similar idea to this post is already applied in the buness world. In my personal homepage, I listed and linked some books from Amazon.com. That makes me write right now. If someone clicks one of my links and buys it from Amazon.com, then it is a reasonable thought that Amazon.com returns a little portion of its profit to the homepage owner who advertises it. This is a sort of the PPP (pay-per-purchase). In my opinion, the personal recommendations in his/her own personal web site may be more strong recommendations than those in Amazon.com by anonymous readers. The profit must be shared with the willing advertisers. By that means, both parties are benefitial. The advertiser (for example, Amazon.com) has (much more) virtual market places out of its own place having strong recommendations and marketing power -- the eager and active Evangelist. The opposite party is trivial. To the end, there must be special type of 'url' such as "http://advertiser/path/show?product=xxx&linkedfrom='bahn'& ...".

I think this idea can be applied to other companies such as google.com and yahoo.com. They can make a advertisement link for a certain 'product', as well as they did for advertisers (company, organization, web site, etc). This is the case that someone types a specific product name in the search field. He/she is possibly willing to buy the item. ... Alas, my poor English...

Don't Make Me an Idiot -- the Creativity Killer.

I found an exciting article from http://www.designtaxi.com/features.jsp (and, http://www.changethis.com) or elsewhere -- "100 Ways to Kill a Concept" originally written by Michael Iva. Here is the full list, but the additional comments by him are available the links above. I'm not responsible for that article (sorry for copying it here without approval).

---
Why Most New Ideas Get Shot Down™

1. The boss won’t go for that.
2. The lawyers won’t go for that.
3. The accountants won’t go for that.
4. The client won’t go for that.
5. The salespeople won’t go for that.
6. The investors won’t go for that.
7. So and so won’t like it.
8. It’s not us.
9. It won’t fit into our system.
10. We’re not ready for that yet.
11. I don’t think it will work.
12. I don’t understand.
13. Do you understand?
14. Will anyone understand?
15. What will they think of next?
16. It's politically incorrect.
17. It’s too complicated.
18. It's too late for that now.
19. It's too expensive.
20. We’ll lose money.
21. Why?
22. What!?
23. Who says?
24. Nobody will want that.
25. Where are you coming from?
26. It can’t be done.
27. Have the committee review it.
28. Let’s do more research.
29. Let’s take a vote on it.
30. Let’s play devil’s advocate.
31. I’ve never heard of such a thing.
32. We’ve already tried that.
33. That’s been done before.
34. It’s not how we do things here.
35. We’ve never done anything like that.
36. We’ll get back to you.
37. That’s a subject for another meeting.
38. It creates more problems than it solves.
39. Sounds like too many problems.
40. That only solves some of the problems.
41. That's going to cause problems.
42. Here we go again.
43. That's unique, but . . .
44. Very interesting, but . . .
45. I understand, but . . .
46. I love it, but . . .
47. Great idea, but . . .
48. Yes, but . . .
49. Sounds simple, but . . .
50. But . . .
51. It’s just not for us.
52. Try again.
53. That sucks.
54. Needs more pizzazz!
55. I don’t like it.
56. Have you thought it through?
57. It won’t sell.
58. Maybe next time.
59. Another day.
60. What are people going to say?
61. People will think we’re nuts!
62. That’s BS.
63. Be realistic.
64. How about this instead?
65. Get a grip!
66. Are you serious?
67. That turns me off!
68. The client is too liberal for that.
69. The client is too conservative for that.
70. What else do you have?
71. So!
72. So what?
73. Oh . . .
74. Oh?
75. Oh, really!
76. Get real, it’s not feasible.
77. Sounds crazy!
78. Nobody does that.
79. Are you kidding me?
80. That’s too off the wall.
81. You can’t argue with success.
82. You can’t fight City Hall.
83. We don’t have time to do it.
84. Let’s not reach beyond our grasp.
85. We’re overextended already.
86. Let’s wait ‘til we see the numbers.
87. You’re proposing what?
88. Let’s not rock the boat.
89. Silence.
90. Laughter.
91. Boos.
92. Where did that come from?
93. I’ve got a better idea.
94. Hey, there’s a recession going on.
95. We’ve never done anything like that.
96. The press will kill us.
97. That’s not your decision to make.
98. We’ll step on too many toes.
99. Just leave it to me; I’ll take care of it.
100. You could lose your job for that . . .
---
Again, check the original source. I'm just a connector to that.

Business & Research Idea: Search Extension

In a previous post, I suggested an idea to cluster search results (either run-time or pre-compiled). Today's thought is derived from that idea, for run-time clustering. Indeed, this is not a clustering approach, but may give the same (or at least similar) results, I guess. I wanted to name this as 'parallel search', but this name is already used for performing the same search in multiple machines.
My idea is to search a keyword with derived terms. Take 'kernel' again for this idea. As mentioned, 'kernel' may mean 'kernel machine' and 'kernel for OS'. So, if one typed 'kernel', then the search engine try to find both searches 'kernel & machine' and 'kernel & OS', as well as 'kernel' itself. The 'Kernel' itself shows the entire search results, but the derived individual may show part of the results for 'kernel'. In other words, the derived/extended search gives one of the clusters for 'kernel'. I think, this idea can also be called as 'parallel search' because we may execute multiple, but different' operations simultaneously.