I started to practice piano on July 10th.  The fact that I’m spending some time to learn it is not coincidental.  A few things can be traced as the root of this newly established interest.

In terms of timing, I have worked very hard for the marketing science conference, and when I came back from Singapore, I spent a whole week working on Bayesian Statistics to prepare for making some progress on that project.  Then I felt very tired, as I spent on-average 15 hours a day to read the books/papers and to program.  Between then and my two big trips to Beijing/Shanghai and Boston, I have about one week.  For this one week, I just want to find something for me to relax.  And just about the right time, the piano for xuanxuan has arrived.

In terms of location, there is nothing more convenient to have a piano in your house.

When I was visiting Jill and Gordon in Chicago, we had a lot of fun singing with Gordon playing the guitar.  Jill is a good pianist, she started piano when she was young, and, of course, was taught to learn classic music.  And Gordon mentioned that it was not hard at all to learn to play the chords on piano and use a guitar fake book to accompany singing.  Since we were buying the piano for xuanxuan, I thought maybe I can learn to sing songs on piano.  

The first thing I learned on piano is the prelude part of “Right Here Waiting”.  Then I learned the whole thing of “Everything I Do, I Do it for You”.  It’s really amazing that these only took me one day to figure out.  Of course, I use very crude ways to play them: I play the melody on the right hand, but once I have the chords added on the left hand, they became beautiful music. 

Then I had some idea, maybe I should learn to play piano, not so ad hoc as I was doing then, but really learn to play well.  This starts my journey of figuring out how to learn piano as fast as possible.

If you have an iPod, and want to keep studying new things, you can actually download MIT lectures and watch them.

Just go to this page. You can find a lot of courses to choose from. This is part of MIT’s opencourseware project.

The following is a screen shot for the course “Differential Equations”.

一级：拜厄学完；二级：车尔尼599后半部；三级：车尔尼849前半部；四级：车尔尼849学完；五级：车尔尼299前半部；六级：车尔尼299后半部；七级：车尔尼299学完及克拉玛；八级：车尔尼740；九级：莫斯科夫斯基练习曲；十级：肖邦练习曲。

钢琴教学可分为启蒙,初级,中级和高级四个阶段,各阶段的学习内容通常包括基本练习,练习曲,复调作品,乐曲几大类.作为基础练习,目前仍然是车尔尼的作品最为系统,如《钢琴初步教程》599,《24首左手练习》718,《钢琴流畅练习曲》849,《钢琴快速练习曲》299,《钢琴练习曲50首》740(699)等.

车尔尼是19世纪上半叶维也纳钢琴演奏学派的代表人物,他所写的大量练习曲是针对古典作品演奏技术,即”古典技术”训练的,主要是手指的颗粒,均匀,快速,对左手的训练较为薄弱,因此期间必须增加各类作曲家的联系曲目作为补充.这些教材可以根据学生的实际情况,有针对性地选择,穿插进行,不能一味的依次一条条的弹下去.

儿童的启蒙阶段

在目前较流行的儿童启蒙教材中大致可分为三种五线谱入门型,即高音谱表入门法,中央c入门法,多音入门法.传统的启蒙教材《拜尔钢琴基本教程》用的就是高音谱表入门法,目前在国内广泛应用的《汤普森浅易钢琴教程》1-5册用的是中央c入门法.

李斐岚,董钢锐编著的《幼儿钢琴教程》采用了以中央c为主的入门法,突破了以往此类教材的音域局限,从3指开始学重量落下,逐步扩大到其余各个手指的方法已被公认为最好的方法.

冯·德·魏尔德:《陶梅格·露丝》系列.这一幼儿系列较注意引起幼儿的兴趣,从小培养他们具有良好的乐感.

柯达依:《钢琴学校》第一册.

巴托克:《小宇宙》.这是一本非常重要的入门课本.因为它的五指位置包括种种非大调或小调结构的五指排列,有助于学生及早接触黑键及各种调式,含有许多复调模仿级对位因素.

齐格勒:《钢琴教本》第一册.该书的特点是先听后弹,耳朵领先.

勋格勒:《钢琴教室》第一册.

狄贝阿里:《二十八首联弹练习曲》中较容易的,越早让学生进行四首联弹训练,越有利于养成用耳朵听辨声部与互相配合的习惯.

钢琴教学的初级阶段

一,钢琴教学初级阶段的练习曲

这阶段的学习内容相当从车尔尼的《钢琴初步教程》599到车尔尼的《钢琴流畅练习曲》849.

1,车尔尼的《钢琴初步教程》599

这是我国钢琴初级阶段教学运用极为广泛的教材之一.全书可分为三个部分.第一部分是57首以前这部分是巩固手型,训练手指独立性的联系,要求弹奏放松,自然,联系速度以慢速中速为主.第二部分是58~79首.58首到70首饰训练快速练习,这里的快速是针对前阶段而言的,在学生走法正确的基础上要求初步加快速度.71首~79首是带装饰音与不带装饰音的旋律练习.练习时先慢速,把节奏弹准确,弹平均,双手对齐,在稍微加快.这部分又是旋律练习要注意乐谱上的表情术语及力度术语,注意分局,呼吸,旋律的连贯与歌唱.第三部分是80首到结束,是综合新的练习,无论音型,节奏型,音阶琶音类和双音练习都比前面丰富,;练习时不仅要弹准确,还要注意音乐起伏,作出力度,速度变化.

2,车尔尼《钢琴简易练习曲》139

这是从599过渡到849的教材之一.139的教材编排不如599集中,可以选择与599不同类型的曲目.在599到50首左右就可以交*使用.

3,车尔尼《24首钢琴左手练习曲》718

这是全部为左手而写的练习,内容有音阶,琶音型,五指型,分解和弦等,学习顺序要灵活掌握,对于手小的学生,可以先学音阶型或五指型的曲目,八度练习暂缓.

4,车尔尼《160首八小节钢琴练习曲》821

本教材不仅技术类型较丰富,而且难度跨度大,包含了从初级程度到高级程度的练习.每一首只有8小节,技术课题非常集中,尤其适合一些其他课业多,练习时间较少的学生使用.

5,莱蒙《钢琴练习曲》37(又译为”勒穆瓦纳”)

全书共50首,由浅入深.他的特点是较多的左手练习,双手比较均衡.可以弥补车尔尼练习曲中偏重右手的不足之处.

6,布格缪勒《简易练习曲25首》100,《练习18首》109

这两集练习曲都是有标题的小曲,形象鲜明,手法简洁,旋律优美动听,音乐性较强,非常适合初学者学习.在599到50首左右时,可以逐步选用作品100的曲目.作品109在技术上比作100难,篇幅也大些.在进入车尔尼的849后可以选用.

7,车尔尼的《流畅练习曲》849

从599到849是初级阶段的前期进入到初级阶段的后期.849共30首,每一首的技术内容基本上不重复.在选用时要具体分析每首练习的难点,考虑技术的”过渡”问题.849与599,139不同之处是每首曲子有了速度标记,根据849教材的标题,应当做到演奏流畅,尽可能按要求的速度去弹,为下一步中级阶段作准备.

I bought a Maxtor Shared Storage Plus external hard drive.

The installation is simple, you just connect it to the network, and you can find it when you install the driver.

There is a small CPU inside the box, and it supports simple Web Server, file sharing capabilities. So I was thinking maybe I can get it to do more things, like mysql, php, etc. Then it will be really working as a linux server. Turned out that someone has worked this out at http://openmss.org

OpenMSS is a replacement firmware for the Maxtor Shared Storage. Currently it’s a modified version of the original firmware which includes:

* The same as the original firmware.
* opt directory pointed to your harddrive for local system files.
* ipkg tool to install your own packages.
* Telnet enabled by default (can easly be disabled and replaced by ssh).

Current experimental packages on ipkg server:

busybox_1.0-8_mipsel.ipk
ctorrent_1.3.4-dnh1-1_mipsel.ipk
dropbear_0.46-4_mipsel.ipk
hddtemp_0.3-beta14_mipsel.ipk
kissd_0.12-1_mipsel.ipk
kissd_0.9-1_mipsel.ipk
lsof_4.75-1_mipsel.ipk
ncftp_3.1.8-1_mipsel.ipk
noip_2.1.1-1_mipsel.ipk
rsync_2.6.6_mipsel.ipk
strace_4.5.11-1_mipsel.ipk
unrar_3.51-1_mipsel.ipk
vsftpd_2.0.3-2_mipsel.ipk

In the discussion forum, there is one how-to to explain how to install MySQL, PHP, HTTP.

This is what a good ad should do.

Each scene refers to one situation matched to the characteristic of a constellation in the zodiac, a total of 12 of them.

Pisces, the fish (February 19 – March 20).

Aries, the ram (March 21 – April 19);

Taurus, the bull (April 20 – May 20);

Gemini, the twins (May 21 – June 21);

Cancer, the crab (June 22 – July 22);

Leo, the lion (July 23 – August 22);

Virgo, the virgin (August 23 – September 22);

Libra, the balance (September 23 – October 23);

Scorpius (also Scorpio), the scorpion (October 24 – November 21);

Sagittarius, the archer (November 22 – December 21);

Capricornus (also Capricorn), the goat (December 22 – January 19);

Aquarius, the water bearer (January 20 – February 18);

In my undergrad ismt101 class, I showed a video called “Did You Know”. Many people liked it. It talks about shifts in education, in technology and in our lifes. I came across a chinese version of that, so let me post both here.

English version:

Chinese version:

The young generation may not appreciate Michael Jackson’s music, but when I was young, his music video was the first I have ever got on hold and really liked. (The sample may not be representative for me as by that time in China, all music are the same, and they are meant to praise something). Anyway, it was refreshing to watch the fatastic dance.

My favorate music of MJ? To me, it was “Heal the World”, I really like the soft melody.

Published on Tuesday 6th of February, 2007 by Online Education Database.
Online search is now a multi-billion dollar industry, with Google alone grossing over $3.5 billion in profits last year. It’s no wonder why so many newcomers are hopping on the search bandwagon, hoping to become the next Google. And those new search engines that may stand the best chance to become the next Google all share one common element — the use of Web 2.0 technology that they hope will increase search result relevance.

Here are 25 such engines. Some offer functionality that’s slowly making its way into traditional search engines. Others further the attempt to traverse the invisible Web and index other previously unsearchable research sources.

Mashups and Tagging

Many of the new search engines use the modular functionality of Web 2.0: mash together several services and add new features.

1. Ajaxwhois.
   Doing a little domain name research? Ajaxwhois takes an existing protocol, WHOIS, and wraps it with a more responsive one. It’s not a traditional search engine per se, but does make finding domain registration information faster. Start typing, and if you stop, it sends out a query. Add a few more characters to the domain name, and the query starts fresh. Results include links to hosting plans, the site (if it’s registered), and Alexaholic, which is a mashup of Alexa, a Web traffic rankings service.
2. FlickrStorm.
   FlickrStorm provides a nice mashup for flickr images. Enter a tag and it comes back with square thumbnails. Scroll through the array, click on images, and they’ll be displayed larger. Add the ones you like to your own “tray”, for later download. It’s a simple but effective interface for consuming photos. An “advanced” feature filters images by license types, including Creative Commons.
3. FundooWeb.
   FundooWeb is a multi-mashup, incorporating results from Yahoo!, Flickr, Yahoo! News, Yahoo! Answers, Amazon, and Yahoo! Maps images. If you search all sources, the results are presented in a couple of formats, including collapsible headlines and a Flickr photo strip, partitioned by source. There’s obviously a heavy leaning to Yahoo, but it’s not a bad way to conveniently compartmentalize several search result sets.
4. Keotag.
   Keotag’s initial face looks quite simple, with font sizes large enough for the dead to read. Type in a keyword or phrase and a line of favicons appear for Google, Technorati, and Bloglines, as well as over a dozen social bookmarking and community news sites. At far left is a Technorati chart showing the number of blog posts containing the key phrase over the past 30 days. Clicking on a particular favicon reveals result headlines for that source, which can be subscribed to through the resulting RSS feed.
5. Whonu.
   Whonu is arguably one of the very first semantic Web search engines available. It offers over 300 search sources and a smart interface that contextualizes what you enter. For example, enter a US ZIP code and whonu presents a set of links to geocode tools including maps, weather maps, and even public events in Google Calendar. There are so many features that the demo screencast video is 26 minutes long. Information is double partitioned by file type and source. The variety of options might be a bit intimidating, but for power research, whonu looks like one of the most promising search tools available, with an effort made to present structured meaning. Killer feature — saved query history using a row of dots.
6. Similicio.us.
   Similicio.us is mashup of del.icio.us which tries to find sites related to a user-entered URL. In other words, “people who liked this site also liked”. This recommendation engine idea is so basic in functionality that it’s a surprise someone didn’t think of it before. The creator of the site admits that similicio.us currently uses shallow searching on del.icio.us to keep queries to their service at a minimum. This is no doubt an engine that could prove useful in other mashups, were it to be extended in scope. (One possibility is to team up with del.icio.us and have access to their full database and engine.)

Rich Internet Application Search Interfaces

The “rich” in RIAs is a matter of personal definition, but engines in this category offer a little something extra in terms of the interface, sometimes employing AJAX.

7. Huckabuck.
   On the surface, Huckabuck seems like any other text search engine, but click on the “search tuner” button and a neat little “equalizer” panel reveals itself. You can use it to give more weight to the different sources — Google, Yahoo, MSN, Technorati, Digg, del.icio.us - as well color-code results, set the slider for results per page, turn on auto-completion on or off, and more. Click on the Presets arrows to reveal predefined equalizations for Research, Shopping, Blog search, Metasearch, Technology Research, and Social Search. Not a bad start to partitioning search results meaningfully.
8. Kartoo.
   Kartoo is yet another search engine that partitions results into several categories, some serious, some frivolous. The presentation, however, is quite different, displayed in little clusters using Flash and icons of a sheet of paper for each result. Some results are more relevant than others, and clicking on an icon takes you to a deeper level of results. This paradigm might be a little confusing at first, but hovering your mouse over a result produces a result summary at left, including a screencap of the result page.
9. KwMap.
   KwMap touts itself as “a keyword map for the whole Internet”. Type in a keyword or phrase, and an unusual interface appears. At right is an alphabetical list of related keyphrases. At left is a visual component showing two axes that resemble an insect’s antennae, dotted with nodes representing related terms. Clicking on a term’s node takes you to another layer of loosely-related terms. This is a new search paradigm, but it offers the opportunity to explore related concepts in small leaps. Thus, a search for the word “tree” could lead you to “tea tree oil” or to a study of ancestor worship (via “family tree”). Hyperlinking mimicks hyper-thought.
10. Mnemomap.
    Mnemomap uses multiple components to display search results. Topmost is a hierarchical graph with nodes branching off the search term. Non-clickable secondary nodes are “Token”, “Tags”, “Translations” and “Synonyms”. Tertiary nodes are search results and can have either a tight relationship to the original search term or a tenuous relationship. Clicking on a tertiary node either adds it to a bar below for a refined search, or produces a new graph, depending on where you click. Below is a section displaying relevant results from Mnemo, Yahoo, flickr, and YouTube. Mnemomap, currently in Alpha 0.2, is a fascinating paradigm for searching, but more suited to power researchers than to the average search engine user.
11. PreFound.
    PreFound, which is powered by Eurekster Swicki, is a simple search engine on the surface, but contains a little slider “equalizer” panel similar to the one in Huckaback (above). PreFound’s panel has settings for music, movies, TV, xBox, etc., instead of search engines. You do have to register to see and use the equalizer (which they oddly call a social search equalizer) but you do not have to ask a question, view previous answers related to your search, or to promote up a search result.
12. Quintura.
    Quintura, who recently received funding, presents text or image search results in a minimalist but graphic form resembling a freeform tag cloud. Holding your mouse cursor long enough over a term in the cloud causes new, related terms to appear in the vicinity of the cursor. While the no click interface is a bit disconcerting at first, you can start over by holding the cursor over the original search term, displayed in red text. Any term in focus (hovered over) generates search results in a scrollable panel below.
13. Ujiko.
    Ujiko has an interface reminiscent of some sort of a video game, presenting results in both a central circle as well as in rows surrounding the circle. The setup allows you to drill down into the categories in the circle or click on actual results on either side, which can be marked as favorites. Ujiko makes a commendable attempt in presenting meaningful results in digestible bites, with a constantly updated interface.
14. Tagnautica.
    Tagnautica starts off with a minimalist interface: a black background and a “CLICK HERE” message. Click and enter your search term, then wait for the strange revolving circle containing numerous spheres on the circumference, which undulate up and down in size. Talk about organic search results. Each result represents a relate term, which can be drilled down into. Or you can click whatever term is in the center (initially the original search term) to get a page of flickr images. Tagnautica is a fascinating photo search parardigm that’s lots of fun and definitely visually inspiring.
15. Topix.
    Ever want to search for topical Web pages and wish you could easily narrow the search to a certain time period? Topix offers just that ability with a neat little interactive timeline map. Clicking on a particular day produces results ordered reverse chronologically from that day backwards. Definitely a handy tool for research, and would be killer mashed up with other functionality.

Social Aspects: User Contribution, Recommendation, Social Networks

Social networks are a hot Web application space, and now they creeping into search engines.

16. Clipfire.
    Michael Arrington of Techcrunch gave Clipfire his blessing, saying how much he likes this ecommerce deal-finding search engine. Sometimes all you need is a simple interface; it’s the members that matter here. The idea is that members submit Web sites, Clipfire searches them, then presents later searchers with product and service deal info. Members are encouraged to use their own affiliate links so that they’re motivated to find good deals and share them. This is a unique idea that’s unlikely to remain so for much longer.
17. Omgili.
    Omgili is a discussion-based engine. In addition to standard search results, a list of links to members is provided who have answered questions relating to a given search term. You can also ask a question, which another member might answer for you with relevant links. Recommendation engines such as omgili have their value in end applications, possibly those similar to the music recommendation site iLike (not to be confused with shopping engine, like).

Visual Search

Engines in this category allow you to search using images and similarity algorithms.

18. Like.
    Like is a “visual shopping” engine that starts off with images of products. Click on an image to get an array of related product images. Use the interface to select a focus area of one image to find similar products by shape or color - say similar sunglasses. Like also lets you filter brands and price ranges. It’s one of the more sophisticated ways to do affiliate marketing. Of course, while you don’t have to enter any text at all to surf’n’shop, the option is there as well.
19. Pixsy.
    Pixsy is a visual search engine for pictures or videos selected from several sources including Buzznet, flickr, iStockphoto, Fotolia, YouTube, and others. Clicking on an image takes you to the source page. For stock photo sites, this might provide copyright and license details. A handy tool for online publishers looking for suitable images to reprint.
20. Retrievr.
    Retrievr is a visual search engine in the truest sense of the term, offering the choice of starting with an image (via URL or uploaded) or a sketch from the user, which can be customized by line thickness and color. Images are then retrieved from flickr. Brilliant concept. The honest truth is that very few of the images in the matrix of results have much resemblance to drawn sketches, but those that do are uncanny. An engine like this is only as good as its algorithms (though it uses brainiac wavelet transforms rather than the traditional neural network algorithms). Still, retrievr is an exciting early- generation advanced search engine offering.
21. Riya.
    Riya visual search, who also offer Like, lets you search amongst people, objects, tags, and photos, as well as gives you a portal to Google, Yahoo, MSN, and flickr. You can browse broadly across the results or drill down through a specific photoset. Results can be emailed, embedded into Myspace or Blogger pages, or subscribed to via the dynamic RSS feed.
22. Tiltomo.
    Tiltomo is yet another flickr mashup that offers a few search options. Enter a single flickr tag or ask for random images. Once you have an array of images, you can find similar images either by theme or by color/ texture. Tiltomo seems to produce slightly more relevant secondary results than some of the other visual search engines.
23. Xcavator.
    Xcavator is another flickr-based engine in its early stages. Currently, it seems a bit limited, as there are only five tags from flickr that can be searched. Selecting one brings up an array of images. Dragging and dropping one of these to the xcavator search box and then selecting a point of interest produces a second, more refined image result set. While these sorts of engines have a ways to go before they’re highly accurate, it’s the promise of what’s to come that’s exciting.

Audio/ Video Search

Up until a few years ago, finding specific music or videos online was a difficult task. Then video search started appearing in traditional search engines. Now, it’s creeping into engines with some advanced features.

24. Liveplasma.
    Liveplasma is a music and video search and discovery engine tied to Amazon.com. Enter an artist, band, movie, director, or actor of interest, and up pops an unusual result set paradigm: floating spheres clustered in overlapping orbits. Each sphere represents information related to the search term. Clicking on a result produces an Amazon summary in the left panel, sometimes with CD/ DVD cover art. Clicking on the summary takes you to its Amazon page. Liveplasma is the type of affiilate marketing search engine that can wag the long tail.
25. Vdoogle.
    Vdoogle is a video search engine that draws its sources from 14 video sharing sites such as YouTube and DailyMotion, as well as veteran sites such as iFilm. Vdoogle is based on Google’s new roll-your-own custom search engine, which is similar to Rollyo. Its Web 2.0 pedigree is tenuous, though it does mashup other Web 2.0 user-contributed services. The accuracy of Vdoogle relies on the proper tagging of source videos, so the engine could do with its own tagging and recommendation engine as an additional layer.

Harvey Danger has just released a new album.  Here is a sample song from the album:

[http://blog.mikezhang.com/wp-content/upload/harveydanger_little_round_mirror.mp3]

I learned about this through a friend who saw this got slashdotted.  I seldom listen to rock music, because I need to concentrate on readings most of the time.  In addition, rock music is not the type for my office.  I don’t want Professor Malone to knock on my door to complain.  But this album is very nice, without the free download, I would have never come across this band. 

In his blog, Jeff Lin, wrote:

Record labels will cling to and fight for old models because that’s all they know (and change is always frightening). But there’s no reason why artists should be limited to just that model. Part of this experiment is to see/demonstrate that doing things in this manner can work, that it is a viable option — both artistically and economically.

I hope they can be successful, not just because of the great music, but also because of the big leap in faith.

I have done some research on music distribution in late 2001/2002, I dropped the project because the technology was not ready for big changes.  I even had a proposal of distributing music through P2P framework.  I can see that some music labels are experimenting some similar ideas.  These are all very encouraging, maybe it is time to look again at this issue now…

 
There are quite a few models out there to experiment “selling” “free” content online.  In 2000, Stephen King released a few chapters of his novel and said if he can get enough payment online (on average $1 per download), he will be releasing the rest part.  He did not get enough money in the end.  Recently, there is an online game design website that released a very well designed game system (a 6-page manual about how to play the game).  They did it because they have received enough donation (about $600, as I can remember), they will be exploring that model in the future for their future releases.  (they call it ransom model http://www.danielsolis.com/meatbot/)

To me, these types of pricing model are destined to fail because people like to free ride.  They give no incentive at all for people to pay the money. 

Harvey Danger’s experiment is different, though.  I certainly believe that there will be people donating money (it is much easier to donate small amount of money now than about 5 years ago).  But I doubt the donation would ever cover the cost ($30,000).  The real value should lie in the word-of-mouth associated with the release of the album. 

In a previous blog entry, I mentioned Steven Wolfram who was the creator of Mathematica. He recently applied his research of “A New Kind of Science“(Book readable online) to art. This is called “Wolfram Tones“, which instantly create unique music in many different genres. After listening to some compositions and creating a few random ones myself, I think it is very cool.

I don’t think it is academically pathbreaking or something like that, but using computer to create art work (or scientific papers in that matter) has been touted by computer scientists for a long long time. This one is actually quite near. If there is a Turing test for music creation, I would give this one a pass.