頁數: 1 2 3 4 5 6 7 - 每頁 20 筆

共有 131 位講者


張智威 (Edward Y. Chang)

張智威 (Edward Y. Chang) Representation Learning on Big and Small Data

The approaches in feature extraction can be divided into two categories: model-centric and data-driven. The model-centric approach relies on human heuristics to develop a computer model to extract features from data. These models were engineered by scientists and then validated via empirical studies. A major shortcoming of the model-centric approach is that unusual circumstances that a model does not take into consideration during its design can render the engineered features less effective. Contrast to the model-centric approach, which dictates representations independent of data, the data-driven approach learns representations from data. Example data-driven algorithms are multilayer perceptron (MLP) and convolutional neural network (CNN), which belong to the general category of neural network and deep learning. In this talk I will first explain why my team at Google in 2006 embarked on the data-drive approach. In 2008, we funded the ImageNet project at Stanford, and subsequently in 2011 filed two data-driven patents, one on data-driven feature extraction and the other on data-driven object recognition. We parallelized five widely used machine learning algorithms including SVMs, PFP, LDA, Spectral Clustering, and CNN, and open-sourced all these algorithms. Recently at HTC, we announced DeepQ Open AI platform, which features these scalable algorithms with enhancements. In particular, I will explain how we have made configuring a distributed system to run CNN both simple and cost effective. In 2012, the world was convinced that the data-driven approach to be effective, after Alexnet achieved breakthrough accuracy in the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) competition. This talk will walk through subsequent enhancements on Alexnet in three perspectives: representation ability, optimization ability, and generalization ability. Unfortunately, most applications face the problem of small data. I will share our experiences with transfer learning and GANs, both positive and negative ones. This talk concludes with a list of open research issues.

張維元 跨界的資料英雄專案-「到院前心肺休止」(OHCA)風險地圖


張傳育 嬰語翻譯機:新生兒哭聲偵測與辨識

新生兒不會說話,只能透過哭聲表達需求。我們研究新生兒的哭聲來判斷嬰兒哭鬧原因,利用大數據及機器學習的方法,分析、判別嬰兒哭聲的各種類型,可隨嬰兒成長修正哭聲模型,並且為每位嬰兒客製化哭聲模型,幫助父母了解嬰兒哭鬧原因。我們與台大醫院雲林分院小兒科醫生合作,收集超過二百萬筆哭聲資料,對於出生四個月內新生兒的哭聲辨識率高達86%,男女嬰兒均適用,收集的有效嬰兒哭聲資料為全球第一。該技術已具體實現,我們開發出全球第一款嬰兒哭聲辨識APP-嬰語翻譯機,「嬰語翻譯機」獲美國路透新聞(Reuters New)、韓國阿里郎新聞網、日本產經新聞、讀賣新聞等超過50個國際媒體報導,具有全球知名度,為嬰兒哭聲辨識技術領先品牌。

張如瑩 語意分析技術的趨勢演變與應用案例分享

近年相當火紅的聊天機器人、虛擬語音助理、智慧客服等技術商品,背後使用的人工智慧(Artificial Intelligence)涵蓋多樣技術領域範疇,其中,為了理解人類的語言意涵,自然語言處理(Natural Language Processing)的語意分析(Semantic Analysis)技術演著舉足輕重的角色。傳統語意分析技術,必須仰賴大量標記語料,為此,必須耗費龐大人力,其所必須耗費的成本,不啻為該技術進行技術商品化過程中相當沈重的負擔。本演講內容將藉由國際大廠的知名案例中所採使用的語意分析技術,引介近年語意分析技術之趨勢演變,包含如何減少語意分析技術研發過程中的人力介入。此外,亦將分享近年新興公司以語意分析技術衍生出的各應用案例。

張宴晟 Real Time Human Body Segmentation on Mobile Device with Deep Learning

Image Segmentation is quite useful for image understanding, however, the computation is considerable as well. In this talk, I'm going to introduce how we utilize deep learning to segment human body, what problems we met and how we resolve it.

張家銘 電腦不只選土豆,還會看病給藥與猜功能?

人類基因草圖在2000年完成後,隨著近來次世代定序技術發展,如:富士康與華大基因 (Beijing Genomics Institute) 將合作生產次世代定序儀 BGISeq-50,代表個人基因體時代即將來臨,如何利用電腦來處理巨量增長的生物資料?演講中將分享幾個應用,首先機器學習在精準醫療,利用個人基因來幫助癌症用藥,機器學習也可以預測蛋白質的功能,這與文件分類的問題很像,但是不同又在哪裡?

張家齊 Quantitative Trading from Zero to One: From Rule-based to Artificial Intelligence

本演講將介紹如何從無到有的構建與設計出一組交易策略。如何從最基本的 Rule-based 的策略,建構出擁有 Machine Learning 或 AI 模式的交易策略與模型。一般來說,交易策略的建構大致上可以分成 Botton-Up 和 Top-Down 兩種,在本次的演講中,我們也將介紹兩種建構模式的差異,以及他們彼此相輔相成的使用方式。希望,能讓大家對自己手上的策略能更了解,也能更知道如何開始下手改善自己手上的策略!

張智傑 理財?機器人?理財機器人?


張智星 音樂檢索與歌聲抽取


張源俊 Learning and Using Statistics Liberally-Exploratory and Elaborate Ways

This is a statistician's view of the development of Data Science. In 2002, Professor Ryan Rifkin, from a mathematical perspective, entitled his PhD thesis as "Everything old is new again." Statistics has a shorter history comparing with mathematics and other natural sciences. In this talk, we will go over some classical statistical papers, and try to connect their thoughts with modern data science. From their thoughts, we wish to relax some barriers as learning and using statistics in such a modern era.

張詠淳 一窺文字探勘之奧妙


曹昱 Machine Learning and Signal Processing for Assistive Hearing and Speaking Devices

With the rapid advancement in speech processing technologies and in-depth understanding of human speech perception mechanism, significant improvement has been made in the design of assistive hearing devices [assistive listening device (ALD), hearing aids (HAs), and cochlear implants (CIs)] to benefit the speech communication for millions of hearing-impaired patients and subsequently enhance their quality of life. However, there are still many technical challenges, such as designing noise-suppression algorithms catered for ALD, HA, and CI users, deriving optimal compression strategies, improving the music appreciation, optimizing speech processing strategies for users speaking tonal languages, to name a few. In this talk, we present our recent research achievements using machine learning and signal processing on improving speech perception abilities for ALD, HA, and CI users.

胡筱薇 社群大數據中的江湖事


崔殷豪 ASAP 比價嗶嗶鳥爬蟲的秘辛與其他電商資料分析案例分享


黃士傑 AlphaGo-深度學習與強化學習的勝利


黃致豪 Calculating WAR of CPBL players 中職球員勝負貢獻值計算

We set a goal to calculate Wins Above Replacement(WAR) players for each player in CPBL, which has never been done before. WAR is widely used in MLB as an index for salary negotiation for free agents. It includes not only the offensive ability of a player, but also the fielding position, fielding ability, and base running ability. Through the project, we hope to show the true value of each player, especially players with good fielding ability, and in turn cause the clubs and the league to appreciate the value of fielding and base running.

黃彥棕 干擾機制對資料科學帶來的挑戰與契機:以抽煙及肺癌存活的因果關係為例

Complexity of the data may provide challenges for making valid causal inference about a scientific hypothesis (smoking vs. lung cancer mortality), particularly in the presence of unknown confounding (underlying lung function). Mendelian randomization (MR) addresses the issue of unknown confounding by using genetic information as an instrumental variable (IV) to estimate the causal effect of an exposure of interest on an outcome. Despite the popularity of IV analyses in fully observable outcomes, methodology is limited for time-to-event survival outcomes with censoring, a common data structure in biomedical sciences. We propose an IV analysis method in the survival context, estimating causal effects on a transformed survival time and the survival probabilities using semiparametric transformation models. We construct unbiased estimating equations to circumvent the difficulty in deriving joint likelihood of the exposure and the outcome, due to the unknown covariance by confounding. Asymptotic properties of the proposed estimators are established. We apply our methods to conduct an MR study for lung cancer survival, which suggests a harmful prognostic effect by smoking intensity (p=0.0067) that would have been missed by the conventional methods.

黃維中 預測式分析於商務服務之應用與挑戰

預測式分析並非新的技術, 過去存在許多統計與經濟模型來預測各種實體跟虛擬世界的行為跟現象。隨著目前資料科技與人工智慧技術的發展, 如何結合龐大的資料來源, 能更有效地進行預測分析, 成為一個有趣且具挑戰的議題。本演講將從整體架構跟實務舉例的方式來跟各愛好者共同探討預測式分析的各種樣貌與未來發展潛力。

黃從仁 認知神經科學 x人工智慧

認知神經科學和人工智慧兩個領域息息相關。認知神經科學研究人類腦神經網路如何賦予我們做物體辨識、語言理解、計畫與決策等認知能力,而人工智慧則透過使用各種類神經網路與演算法來達成如人般的認知歷程。如原本是認知神經科學家的DeepMind執行長Demis Hassabis博士所言:認知神經科學能對人工智慧能有突破性的啟發。反之,人工智慧也對理解人類大腦如何運作有革命性的影響。本場演講將藉由一些實例討論這兩個領域如何互利互用,並鼓勵雙方能有更緊密的交流與互動。
頁數: 1 2 3 4 5 6 7 - 每頁 20 筆