5/18/2015

台大研發軟體 提高甲狀腺檢測準確度

洪欣慈,台大研發軟體 提高甲狀腺檢測準確度,中國時報,2015 年 05 月 18 日

 
甲狀腺結節電腦輔助偵測診斷軟體可將超音波影像彩色視覺化、標示腫瘤輪廓,幫助醫師判讀。(洪欣慈攝)

Facebook 產品分析師的面試問題

根據 Jewel 在 Inside 的報導 
Facebook 產品分析面試問題:「在一條長度為 1 的直線上隨機取兩點,由這兩點分成的三個線段能形成一個三角形的機率是多少?」 
"Two points are randomly placed on a line of length 1. What is the probability that the three segments created form a triangle?” — Facebook Product Analyst

工業胃散

何欣潔,工業胃散 砸了台灣國際製藥信譽,今周刊,955期,2015/04/09
知名廠牌胃散、化咳散使用工業用原料,讓台灣加入「全球最嚴謹製藥規範 PIC/S 認證的努力破功,國際上對台灣的製藥品質信心也大打折扣。部分老牌藥廠面臨轉型壓力的態度消極,衛福部稽查、輔導不力,是藥安問題爆發主因。 
多家知名廠牌胃散、化咳散使用工業用原料碳酸鎂或碳酸鈣,不但可能傷害病人健康,甚至可能傷害台灣藥品在世界上的商業信譽。 
台灣自二○一三年元旦開始,正式成為國際醫藥品稽查協約組織(PIC/S)會員國,並規定自一五年元旦開始,國內藥廠必須符合 PIC/S 製造標準,否則不得繼續製藥。未料政策上路三個月,就出現嚴重的藥安危機。 
PIC/S認證,是目前全球公認最嚴謹的製藥規範。一九七○年由歐盟國家發起,至今共有四十八個會員國,通過認證代表製藥水準能與世界接軌。依據歐盟一一年最新修訂之法規規定,自一三年七月起,輸入歐盟之原料藥須符合歐盟認可的藥品規範,PIC/S 便是其中之一。... 
為何全球最嚴謹的製藥規範,到台灣卻變得荒腔走板?風險管理專家、台大公衛系教授吳焜裕分析,台灣許多食品、藥品廠規模都小,且是家族企業,要建立完整的內部原料管理機制,相當困難。 
因推動 PIC/S 認證,須耗費上億元投資改善藥廠設備,業界原本預測,在正常市場機制下,藥廠將出現購併潮,會有一半以上的藥廠退出市場。但在這次事件中正長生、木村的案例來看,部分未通過認證藥廠決定改走「委外代工」路線,或以工業用原料來節省成本,PIC/S 所預期的提升用藥安全,顯然大打折扣。

5/16/2015

Convex Optimization (凸函數最佳化)

由於內點演算法 (interior point algorithm) 的進步,凸函數最佳化變成解決問題的重要工具。

Giuseppe C. Calafiore and Laurent El Ghaoui, Optimization Models, Cambridge University Press, 2014.

此書分成三大部份 (linear algebra models, convex optimization models, applications),有許多令人驚奇且有趣的應用,適合做為大學部高年級和碩士班課程的教科書。

Stephen Boyd and Lieven Vandenberghe, Convex Optimization, Cambridge University Press, 2004.

則是另一本風格接近的書,適合做為碩博士班課程的教科書。

5/14/2015

Top Equations in the (Control) Field

J.P. How, Top Equations in the Field, IEEE Control Systems Magazine, April 2015.

模仿 Ian Stewart 的書,作者提出了控制工程中重要的方程式,包含

Lyapunov’s Stability Theorem
Sensitivity Constraints
Lagrange Multipliers
Optimal Control (The Hamilton–Jacobi–Bellman (HJB) equation, LQG)
Riccati Equation (LQR, H inifinity)
Bayes’ Rule
Least Squares (Kalman filter)
Small Gain (structured singular value)
S-Procedure (LMI)

Seventeen Equations that Changed the World

Ian Stewart, Seventeen Equations that Changed the World, Profile Books, 2013.

Larry Phillips 畫的圖

嵌入永久的圖片連結
Andy Kiersz 有進一步的說明

5/06/2015

Strategy Rules

David B. Yoffie and Michael A. Cusumano, Strategy Rules: Five Timeless Lessons from Bill Gates, Andy Grove, and Steve Jobs, HarperBusiness, 2015.

本書說明其成功的五大方法,書商的說明如下
  1. Look Forward, Reason Back: They determined where they want their companies to be in the future and could "reason back" to identify the moves that would take them there.
  2. Make Big Bets, Without Betting the Company: All three men made enormous strategic bets but rarely took gambles that put the financial viability of their companies at undue risk.
  3. Build Platforms and Ecosystems: Technology leaders have to create industry platforms that enable other firms to create complementary products and services that make the platforms increasingly valuable.
  4. Exploit Leverage and Power: Gates, Grove, and Jobs often turned opponents’ strengths into weaknesses and used enormous resources (once they had them) to dominate competitors.
  5. Shape the Company around Your Personal Anchor: From Gates’ understanding of software to Grove’s devotion to process discipline and Jobs’ obsession with design, all three built their companies around their personal strengths while compensating for their weaknesses.
方法一是動態規劃 (Dynamic programming) 的概念,賽局理論 (Game theory) 只是應用此理論。不過,困難的地方在於高度不確定的環境中,如何看出未來的走向。