預備研究生規定 (資訊處，新增加同學的問題)
Antonio Torralba, Phillip Isola, and William Freeman, Foundations of Computer Vision, The MIT Press, 2024.
作者是三位 MIT 的教授。如果你想了解電腦視覺 (Computer vision) 相關研究，甚至機器學習、研究方法，大力推荐。我有空就翻一下，增加對整個領域的了解，每次都有驚喜，很神奇的書。例如
(4/9/2025) 許志華 (The original speaker needs to take care of emergency affairs.)
Robin Wigglesworth, Daron Acemoglu is not having all this AI hype, Financial Times, May 28 2024.
Who is Daron Acemoglu? 246,645 Citations so far!
AlphaProof and AlphaGeometry teams, AI achieves silver-medal standard solving International Mathematical Olympiad problems, 25 JULY 2024.
鉅亨網新聞中心，台積電董事長劉德音撰文 談1兆電晶體GPU的到來，2024-03-29
文中指出，從 1997 年擊敗西洋棋人類冠軍的「深藍」，到 2023 年爆火的 ChatGPT，再過 15 年，人工智慧已經發展到可以「合成知識」(synthesize knowledge) 的地步，可以創作詩歌、創作藝術品、診斷疾病、編寫總結報告和電腦程式碼，甚至可以設計與人類製造的積體電路相媲美的積體電路。
為了提高同學們的學習動機，提供以下相關的資訊，以幫助同學們找到方向。也和暑期實習和未來就業中，決策支援系統中的演算法有密切關聯。以下許多的內容屬於碩博士階段的課程，也可以增加同學們就讀研究所的動機：
Rui Wang, Danielle Robinson, Christos Faloutsos, Yuyang Wang, and Rose Yu, AutoODE: Bridging physics-based and data-driven modeling for COVID-19 forecasting, NeurIPS 2020 Workshop on Machine Learning in Public Health. (best paper award at the NeurIPS Machine Learning in Public Health Workshop)
As COVID-19 continues to spread, accurately forecasting the number of newly infected, removed and death cases has become a crucial task in public health. While mechanics compartment models are widely-used in epidemic modeling, data-driven models are emerging for disease forecasting. In this work, we investigate these two types of methods for COVID-19 forecasting. Through a comprehensive study, we find that data-driven models outperform physics-based models on the number of death cases prediction. Meanwhile, physics-based models have superior performances in predicting the number of infected and removed cases. In addition, we present an hybrid approach, AutoODE, that obtains a 57.4% reduction in mean absolute errors of the 7-day ahead COVID-19 trajectories prediction compared with the best deep learning competitor.
Dirk P. Kroese, Zdravko Botev, Thomas Taimre, and Radislav Vaisman, Data Science and Machine Learning: Mathematical and Statistical Methods, Chapman & Hall, 2019. (Lecture notes and Python files)
The purpose of Data Science and Machine Learning: Mathematical and Statistical Methods is to provide an accessible, yet comprehensive textbook intended for students interested in gaining a better understanding of the mathematics and statistics that underpin the rich variety of ideas and machine learning algorithms in data science.
Cynthia Rudin, Chaofan Chen, Zhi Chen, Haiyang Huang, Lesia Semenova, and Chudi Zhong, Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges. Statistics Surveys, 2022.
Interpretability in machine learning (ML) is crucial for high stakes decisions and troubleshooting. In this work, we provide fundamental principles for interpretable ML, and dispel common misunderstandings that dilute the importance of this crucial topic. We also identify 10 technical challenge areas in interpretable machine learning and provide history and background on each problem. Some of these problems are classically important, and some are recent problems that have arisen in the last few years. These problems are: (1) Optimizing sparse logical models such as decision trees; (2) Optimization of scoring systems; (3) Placing constraints into generalized additive models to encourage sparsity and better interpretability; (4) Modern case-based reasoning, including neural networks and matching for causal inference; (5) Complete supervised disentanglement of neural networks; (6) Complete or even partial unsupervised disentanglement of neural networks; (7) Dimensionality reduction for data visualization; (8) Machine learning models that can incorporate physics and other generative or causal constraints; (9) Characterization of the “Rashomon set” of good models; and (10) Interpretable reinforcement learning. This survey is suitable as a starting point for statisticians and computer scientists interested in working in interpretable machine learning.
作者／張紹敏 圖片來源：卓杜信，麥當勞、美芝城都是他客戶！七年級面板主管回大武山「賣雞蛋」，靠數據幫爸爸重振家業，Cheers：快樂工作人，2022/08/13
通路
為什麼要自己賣蛋？魏毓恆解釋：「傳統蛋商只要缺蛋，每天都會開車在出貨碼頭等你，不讓你把蛋給別人；可是蛋很多的時候，他們的貨車就會『連續壞一個禮拜』，遲遲不來。」產銷的不平衡，成為大武山陷入困境一大原因，而魏毓恆不打算視而不見。
李忠謙，他高一輟學寫詩、大學6年才畢業，他是今年的菲爾茲獎得主：39歲的普林斯頓數學教授許埈珥，風傳媒，2022-07-06
相較於他在少年時期厭惡抵抗數學（據稱是小學一次數學考試考差所致）、甚至高中一度輟學寫詩的迷茫過去，許埈珥的人生轉折確實比在麻省理工學院拖地擦黑板的叛逆威爾還要傳奇。...
Bobak Kiani, Randall Balestriero, Yann Lecun, and Seth Lloyd, projUNN: efficient method for training deep networks with unitary matrices, arXiv:2203.05483v2.
In learning with recurrent or very deep feed-forward networks, employing unitary matrices in each layer can be very effective at maintaining long-range stability. However, restricting network parameters to be unitary typically comes at the cost of expensive parameterizations or increased training runtime. We propose instead an efficient method based on rank-k updates -- or their rank-k approximation -- that maintains performance at a nearly optimal training runtime. We introduce two variants of this method, named Direct (projUNN-D) and Tangent (projUNN-T) projected Unitary Neural Networks, that can parameterize full N-dimensional unitary or orthogonal matrices with a training runtime scaling as O(kN^2). Our method either projects low-rank gradients onto the closest unitary matrix (projUNN-T) or transports unitary matrices in the direction of the low-rank gradient (projUNN-D). Even in the fastest setting (k=1), projUNN is able to train a model's unitary parameters to reach comparable performances against baseline implementations. By integrating our projUNN algorithm into both recurrent and convolutional neural networks, our models can closely match or exceed benchmarked results from state-of-the-art algorithms.
Mordechai Rorvig, Computer Scientists Prove Why Bigger Neural Networks Do Better, Quanta Magazine, February 10, 2022.
Interpolation
An old mathematical result says that to fit n data points with a curve, you need a function with n parameters. (In the previous example, the two points were described by a curve with two parameters.) When neural networks first emerged as a force in the 1980s, it made sense to think the same thing. They should only need n parameters to fit n data points — regardless of the dimension of the data.
“This is no longer what’s happening,” said Alex Dimakis of the University of Texas, Austin. “Right now, we are routinely creating neural networks that have a number of parameters more than the number of training samples. This says that the books have to be rewritten.”
