講題：
COMMON CAVEATS OF MACHINE LEARNING
講者：
張智星, Jyh-Shing Roger Jang
/ 科技長 (CTO), 玉山金控 (E.SUN Financial Holding Company)
講者簡介：
張智星於1992年獲得美國加州大學柏克萊分校電機電腦博士，博士論文為類神經網路與模糊邏輯之建模與應用。1993~1995任職於美國麻州 MathWorks 公司，負責開發與 MATLAB 合用的Fuzzy Logic Toolbox。1995年起回台任教於清華大學資訊系，2012年轉任台大資訊系。2017-2019年擔任台大醫院資訊室主任，2018年起擔任台大金融科技研究中心主任，2020年借調到玉山金控擔任科技長。專長在於機器學習之各項應用，包含語音辨識、音樂檢索、文件分類、影像辨識、醫療數據與金融資料分析等領域。
https://udn.com/news/story/7239/4764757
講題摘要：
Instead of introducing the newest technologies in ML, this talk will pull us back and review some of the common caveats when we are in the routines of ML practice. Most of these caveats occur when the dataset is small or imbalanced, or when the dimension is large. Moreover, most of the caveats can be detected in advance by exerting more caution in coding. We will give some examples of how these caveats might occurs and how to prevent them . By knowing these caveats, you will be able to have a complete view of ML without falling into the fallacy of it.
