Neural Network Project
作者:wbc 日期:2009-12-12
上了一個學期的CSC3230,對於AI有了初步瞭解,前期課程比較悶,入肉的是最後的Neural Network,終於見到AI的強大。
學期完了,但還要交project,這些project也費事找group mate了,自己單拖,可惜對於NN的掌握還不太熟悉,tutor又無去過,做起來非常迷茫,通過分析那字字珠璣的project spec. 慢慢摸著石頭過河,也有了少少進展。
今日又在google搵到篇非常簡單易明的Backward Propagation的文章,興奮不已,必須收藏。
ref. http://clemens.bytehammer.com/papers/BackProp/index.html
引用內容
Backward Propagation Tutorial
Introduction
Backward Propagation is probably the most common method for
training forward-feed neural networks. A forward pass using an input pattern
propagates through the network and produces an actual output. The backward
pass uses the desired outputs corresponding to the input pattern and updates
the weights according to the error signal.
There are hundreds of papers covering the subject of backward
propagation. Unfortunately, many of them tend to exhibit a vast stockpile
of equations and complicated partial derivatives with undefined variables
to explain a concept that is really quite simple. Quite often, a pseudocode
algorithm or an example with pictures is the most effiecient method to convey
information.
Artificial Neuron Anatomy
請稍等,圖片正在下載中...
w --> weight
I --> input
O --> output
A --> activation value
Think of the activations as a pulses of electricity traveling
along the dendrite path. Each activation pulse is multiplied by the weight
of the dendrite it is traveling along. When all inputs to the neuron have
been "stimulated" with this pulse, the neuron adds up these products
and performs a thresholding or "squashing" operation on the resulting
sum. The result of the squashing fun...
學期完了,但還要交project,這些project也費事找group mate了,自己單拖,可惜對於NN的掌握還不太熟悉,tutor又無去過,做起來非常迷茫,通過分析那字字珠璣的project spec. 慢慢摸著石頭過河,也有了少少進展。
今日又在google搵到篇非常簡單易明的Backward Propagation的文章,興奮不已,必須收藏。
ref. http://clemens.bytehammer.com/papers/BackProp/index.html
引用內容Backward Propagation Tutorial
Introduction
Backward Propagation is probably the most common method for
training forward-feed neural networks. A forward pass using an input pattern
propagates through the network and produces an actual output. The backward
pass uses the desired outputs corresponding to the input pattern and updates
the weights according to the error signal.
There are hundreds of papers covering the subject of backward
propagation. Unfortunately, many of them tend to exhibit a vast stockpile
of equations and complicated partial derivatives with undefined variables
to explain a concept that is really quite simple. Quite often, a pseudocode
algorithm or an example with pictures is the most effiecient method to convey
information.
Artificial Neuron Anatomy
請稍等,圖片正在下載中...
w --> weight
I --> input
O --> output
A --> activation value
Think of the activations as a pulses of electricity traveling
along the dendrite path. Each activation pulse is multiplied by the weight
of the dendrite it is traveling along. When all inputs to the neuron have
been "stimulated" with this pulse, the neuron adds up these products
and performs a thresholding or "squashing" operation on the resulting
sum. The result of the squashing fun...
Tags: AI neural network
Image resize
作者:wbc 日期:2009-10-17
Blue Screen casued by afd.sys
作者:wbc 日期:2009-10-02
Visual Studio 2008 無法卸載
作者:wbc 日期:2009-09-28
From http://coolwanglu.blogspot.com/2008/09/visual-studio-2008.html
引用內容
引用內容症狀是用Control Panel的uninstall/repair,或者源安裝盤,都在load某個component時掛掉,不能進入正常安裝界面,然後彈出對話框,提示“A Problem has been encountered while loading the setup components. Canceling setup.”
而Log中顯示錯誤
MSITOSIT: [2] ERROR processed; exception was thrown for retail build
Developer Comment: : has no associated directory, verify the Feature table is correct
Build Time: Thu Dec 11 14:04:18 2003
HRESULT: -2147467259
網上查了一下,有的說是SP1的問題,有的說是Kaspersky的問題,我這裏都不管用。
最後的解決辦法是找到源安裝盤的vs_setup.msi,用Admin進入cmd,運行msiexec /x vs_setup.msi,它似乎就什麼都不檢查,就能卸載了。
之後重裝也沒問題。
而Log中顯示錯誤
MSITOSIT: [2] ERROR processed; exception was thrown for retail build
Developer Comment: : has no associated directory, verify the Feature table is correct
Build Time: Thu Dec 11 14:04:18 2003
HRESULT: -2147467259
網上查了一下,有的說是SP1的問題,有的說是Kaspersky的問題,我這裏都不管用。
最後的解決辦法是找到源安裝盤的vs_setup.msi,用Admin進入cmd,運行msiexec /x vs_setup.msi,它似乎就什麼都不檢查,就能卸載了。
之後重裝也沒問題。
Google Apps Profiles API Client
作者:wbc 日期:2009-07-29
Reply From Google API team:
引用內容
My output...
請稍等,圖片正在下載中...
引用內容To add additional information to domain users it is necessary to use the Profiles API. Since the release of this API, the Shared Contacts API should be use only for external contacts. For more information see:
http://code.google.com/apis/apps/profiles/developers_guide_protocol.html
http://code.google.com/apis/apps/profiles/developers_guide_protocol.html
My output...
請稍等,圖片正在下載中...
Google Shared Contacts Manager
作者:wbc 日期:2009-07-20
核心代碼來自:http://code.google.com/p/google-shared-contacts-client/
想不到google gmail的企業級應用到現在還沒有正式出share contact manager,只是開放了部份API。
在web上面根本沒有任何管理UI。
API也好,起碼還可以管理。但有一個更不便的是,API做出的操作,不會立即在web內見到效果。據Google說效果會在24小時內可見。
如果等不急官方版shared contact manger,或許可以試試看我這個:
http://www.youtube.com/watch?v=UYsAUUeNEQw
VPN in Windows 7
作者:wbc 日期:2009-07-05
Barcode Generator Add-in for Word & Excel
作者:wbc 日期:2009-04-21
這兩天心血來潮,打算試試看office的add-in開發.
就用之前搞定的barcode generator來繼續開發。
當然,目前依然只支持 code 128,code39,code93
效果如下:
在微軟的網上看了些文章及i示例,
點擊下載此文件,詳細示範了Office 2007 新加入的Ribbon UI使用方法。
更多資料:http://msdn.microsoft.com/en-us/library/aa434078.aspx
昨天搞定了Excel的調用,今天則繼續做word的。
兩個程序都遇到不同的麻煩事:
Excel插入圖片的座標計算:
Pixel to Point,
Point to Inch,
Inch to #Character
Point to #Character
...非常麻煩.
圖像長寬的計算:
就用之前搞定的barcode generator來繼續開發。
當然,目前依然只支持 code 128,code39,code93
效果如下:
在微軟的網上看了些文章及i示例,
點擊下載此文件,詳細示範了Office 2007 新加入的Ribbon UI使用方法。更多資料:http://msdn.microsoft.com/en-us/library/aa434078.aspx
昨天搞定了Excel的調用,今天則繼續做word的。
兩個程序都遇到不同的麻煩事:
Excel插入圖片的座標計算:
Pixel to Point,
Point to Inch,
Inch to #Character
Point to #Character
...非常麻煩.
圖像長寬的計算:
Database 損毀
作者:wbc 日期:2009-04-17
使用了access 2007新支援的database(accdb),一個多月來都運作正常,這幾天,出現了問題,在server里,asp可以正確read write 該database,但我下載下載,用access打開,提示資料庫損毀,運氣好還能看到tablename,否則,一開該database,access 就死。
因為有時能開啟,想趁這個機會,把資料庫壓縮輸出,或者另存...完全失敗。
又嘗試過開一個新的資料庫,然後導入這資料庫的結構....失敗。
由於table不多,於是嘗試讓asp讀出已有的field name,然後人手抄一份吧。
奇怪的是,已死的database雖然只能看到表名,不能看到內部數據,但卻可以右鍵複製數據,然後粘貼去新的database.
不管怎麼說,終於還原了....得自動備份才行...
要是數據大量,看來直接在asp里讀取結構再建立更快..
請稍等,圖片正在下載中...
因為有時能開啟,想趁這個機會,把資料庫壓縮輸出,或者另存...完全失敗。
又嘗試過開一個新的資料庫,然後導入這資料庫的結構....失敗。
由於table不多,於是嘗試讓asp讀出已有的field name,然後人手抄一份吧。
奇怪的是,已死的database雖然只能看到表名,不能看到內部數據,但卻可以右鍵複製數據,然後粘貼去新的database.
不管怎麼說,終於還原了....得自動備份才行...
要是數據大量,看來直接在asp里讀取結構再建立更快..
請稍等,圖片正在下載中...
幻燈片演示 : 








