Newsletter No. 483

財 金 淺 趣 / F inancially F riendly 中大人工智能及深度學習研究成績斐然 Faculty of Engineering’s Achievements in Artificial Intelligence and Deep Learning 安心寧神專家言 The Placebo Effect of Expert Opinions 牛熊博士是資深證券經紀行從業員,在電臺主持財經節目,歡迎聽眾致電諮詢股票投資心 得。以下是兩段典型對話。 甲女士: 我想請問你對5號(滙控)的看法。 牛博士: 甚麽價位買入的? 甲女士: 2012年底七十元的時候。 牛博士: 那麽到現在收息也應有十五元左右了,賺了不少呢! 甲女士: 上星期它又升了差不多半成,那該怎樣做才好呢? 牛博士: 你可再持一陣子。如果深港通計劃比預期早實行,股價很容易會再升一成的,那你 便可賺得更多。 乙先生: 請問你對滙控怎麽看? 牛博士: 甚麽價位買入的? 乙先生: 四十元。 牛博士: 嘩,這個價位挺不錯呢,定是2009年全球金融風暴後入的貨吧。 乙先生: 那現在該怎麽辦? 牛博士: 滙控因為推出回購計劃,加上預期實行深港通計劃,這幾天股價升了不少。不過, 如果回購提早結束,或者深港通短期內還未開通,股價可能輕易跌個一成。你可考慮沽出套 現,不冒這個險。 專家問來電聽眾「甚麽價位買入?」,平常不過,聽起來買入價好像事關重大。事實上,那與 一隻股票未來的價格並無關係。無論是股票,或任何投資工具,買、沽或持的決定都不跟買 入價掛鈎。這是「歷史價格無關前景」的概念。滙豐或者任何一隻股票日後股價如何,跟你 是在2012年用七十元或2009年用四十元買入,一點關係都沒有。 節目主持人與來電者的對話通常都是廢話。即是說,那些所謂忠告根本毫無價值也不專業。 說到底,這些電臺節目也有一項重要功用:大部分致電者都想聽給他們安慰的「好話」,而 並非指點他們投資的「專家意見」。這樣看來,節目主持人無疑是給焦慮的投資者開了一方 安慰劑。 Dr. Niu Xiong is a veteran in the stock brokerage business and hosts a radio programme for audience to phone in to get some advice on their stock investment. Here are two typical conversations. Ms. A: I want to hear your opinion on ‘0005’ (HSBC). Dr. Niu: How much did you pay for it? Ms. A: I bought it at $70 in late 2012. Dr. Niu: Then you should have collected about $15 dividend. You have now made good profits. Ms. A: What should I do with it given it has risen more than 5% in the last week? Dr. Niu: You can hold on to the stock for a while. If the Shenzhen-Hong Kong stock connect scheme is implemented sooner than expected, then the price can easily go up another 10%. Then you may improve your profits further. Mr. B: What’s your view on HSBC? Dr. Niu: How much did you pay for it? Mr. B: I bought it at $40. Dr. Niu: Wow, it was a very good price. You must have bought it in 2009 after the Global Financial Crisis. Mr. B: What should I do now? Dr. Niu: HSBC has risen quite a lot in the past few days in response to the company repurchase programme and the anticipated Shenzhen-Hong Kong stock connect scheme. However, if the repurchase programme ends soon or the Shenzhen-Hong Kong thing does not happen soon, then the stock price could easily drop by 10%. You can consider cashing in now and not take the risk. It is very common for a pundit to begin by asking the caller, ‘How much did you pay for it?’ as if the purchase price really matters. In fact, it has nothing to do with the future price of a stock. For any stock or other instrument of investment, the decision to buy, hold or sell is independent of how much an investor has paid for it. This is the notion of ‘Historical Price is Irrelevant’. The future price of HSBC, or any stock for that matter, has nothing to do with whether you bought it at $70 in 2012 or at $40 in 2009. The conversations between the host and the callers are generally nonsense, i.e., there is no meat and nothing expert about the advice given. But these radio programmes do serve one important function: the majority of callers need some ‘good words’ to comfort them rather than ‘expert opinions’ to guide them in investing. In this sense, they prescribe the placebos for the anxious investors. 由中大電子工程學系歐陽萬里教授(上左)、李鴻升教授(上右)和王曉剛教授(上中)帶領 的CUVideo團隊,早前參加被視為當代電腦視覺技術奧林匹克賽的「ImageNet大規模視 覺識別挑戰賽」,取得佳績。大賽共有四個比賽項目,即圖像物體檢測、圖像物體定位、視 頻物體檢測及場景分類,其中「視頻物體檢測」項目的參賽者需要從上千個視頻中,檢測三 十類物體在視頻中的準確位置。CUVideo團隊成功從三十項指定檢測項目中,取得二十八 項優勝,平均檢測率高達67.8%,其檢測數量及檢測準確率的成績更達至世界第一。 此外,中大多媒體實驗室最近在NVIDIA於美國矽谷舉行的圖形處理器(GPU)科技年度 大會上,被譽為「人工智慧研究先鋒」,與美國加州大學柏克萊分校、麻省理工學院、 史丹福大學及牛津大學等齊名,是亞洲唯一獲此美譽的高等學府。 中大多媒體實驗室於2001年由信息工程學系湯曉鷗教授創辦,是本港最早應用深度學 習於電腦視覺研究的團隊。NVIDIA是GPU的發明者,亦是領先全球的視覺運算技術廠 商,其主辦的年度大會雲集頂尖業界開發者及研究人員。中大多媒體實驗室可優先使用 NVIDIA在深度學習的硬體及軟件,以加強研究設備及技術支援,包括視頻監控、網絡規 模影像及錄像搜索,以及人機交互系統等,進一步推動全球深度學習和電腦視覺技術的 發展。 Under the leadership of Prof. Ouyang Wanli (above left), Prof. Li Hongsheng (above right) and Prof. Wang Xiaogang (above centre) of the Department of Electronic Engineering, the CUVideo team has won in the ‘ImageNet Large Scale Visual Recognition Challenge’ (ILSVRC), one of the major competitions in computer vision. There were four categories in the challenge, namely, objection detection, object localization, object detection from video, and scene classification. The ‘object detection’ challenge requires each team to detect 30 object categories from 1,000 videos. CUVideo team has successfully detected 28 object categories in this challenge with the mean Averaged Precision (mAP) of 67.8%—the best result ever achieved by competitors of this challenge. The CUHK Multimedia Laboratory has also been recently recognized as one of the pioneers in AI research at the GPU Technology Conference 2016 hosted by NVIDIA in Silicon Valley. CUHK has become the only institution in Asia to receive this recognition, alongside the University of California, Berkeley, Massachusetts Institute of Technology, Stanford University and the University of Oxford. The CUHK Multimedia Laboratory was established by Prof. Tang Xiaoou of the Department of Information Engineering in 2001. It was the first research team that applies Deep Learning to Computer Vision in Hong Kong. CUHK will have priority pre-release access to hardware and software provided by NVIDIA to advance the University’s research facilities and technical support for video surveillance, web scale image and video search, as well as human and computer interaction. 06 # 4 8 3 | 1 9 . 0 9 . 2 0 1 6

RkJQdWJsaXNoZXIy NDE2NjYz