1/31/2021

I Know What You Bought At Chipotle for $9.81 by Solving A Linear Inverse Problem

Michael  Fleder and Devavrat D. Shah, I Know What You Bought At Chipotle for $9.81 by Solving A Linear Inverse Problem, Proceedings of the ACM on Measurement and Analysis of Computing Systems, November 2020, Article No. 47.  

We consider the question of identifying which set of products are purchased and at what prices in a given transaction by observing only the total amount spent in the transaction, and nothing more. The ability to solve such an inverse problem can lead to refined information about consumer spending by simply observing anonymized credit card transactions data. Indeed, when considered in isolation, it is impossible to identify the products purchased and their prices from a given transaction just based on the transaction total. However, given a large number of transactions, there may be a hope.

1/25/2021

Special Issue — M&SOM 20th Anniversary

Special Issue — M&SOM 20th Anniversary, Volume 22, Issue 1, January-February 2020 (online)

This special issue contains invited and review articles by eminent researchers in the field.

1/24/2021

Data-Driven Modeling and Optimization of the Order Consolidation Problem in E-Warehousing

Fatma Gzara, Samir Elhedhli, Ugur Yildiz, and Gohram Baloch, Data-Driven Modeling and Optimization of the Order Consolidation Problem in E-Warehousing,  INFORMS Journal on Optimization, Vol. 2, No. 4, Fall 2020, pp. 273–296. (online pdf)

We analyze data emanating from a major e-commerce warehouse and provided by a third-party warehouse logistics management company to replicate flow diagrams, assess order fulfillment efficiency, identify bottlenecks, and suggest improvement strategies. Without access to actual layouts and process-flow diagrams and purely based on data, we are able to describe the processes in detail and prescribe changes. By investigating the characteristics of orders, the wave-sorting operation, and the order-preparation process, we find that products from different orders are picked in batches for efficiency. Similar products are picked in small containers called totes. Totes are then stored in a buffer area and routed to be emptied of their contents at induction lines. Orders are then consolidated at the put wall, where each order is accumulated in a cubby. This order consolidation process depends on the sequence in which totes are processed and has a huge impact on order-completion time. We, therefore, present a generalization of the parallel machine–scheduling problem that we call the order consolidation problem to determine the tote-processing sequence that minimizes total order completion time. We provide mathematical formulations and devise heuristic and exact solution methods. We propose a fast simulated annealing metaheuristic and a branch-and-price approach in which the subproblems are variants of the single machine-scheduling problem and are solved using dynamic programming. We also devise a new branching rule, compare it against the literature, and test it on randomly generated and industry data. Applied to the data and the warehouse under study, optimizing the order consolidation is found to decrease the completion time of 75.66% of orders and achieve average improvements of up to 28.77% in order consolidation time and 21.92% in cubby usage.

1/23/2021

計算機程式補考後的人生 (2/2):對教學的影響

因為自己失敗的經驗,所以課程中特別強調實作和電腦軟體的使用。

首先,我的講義中,寫下使用軟體的步驟,並請學生於學期初影印所有的上課筆記;課堂中,使用放大鏡或是 Excel 下檢視 200%,讓最後一排的學生也看得清楚,所有的展示或步驟一定操作兩遍。

1/22/2021

Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead

C. Rudin, Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead, Nature Machine Intelligence, 1, 206–215 (2019). 

Black box machine learning models are currently being used for high-stakes decision making throughout society, causing problems in healthcare, criminal justice and other domains. Some people hope that creating methods for explaining these black box models will alleviate some of the problems, but trying to explain black box models, rather than creating models that are interpretable in the first place, is likely to perpetuate bad practice and can potentially cause great harm to society. The way forward is to design models that are inherently interpretable. This Perspective clarifies the chasm between explaining black boxes and using inherently interpretable models, outlines several key reasons why explainable black boxes should be avoided in high-stakes decisions, identifies challenges to interpretable machine learning, and provides several example applications where interpretable models could potentially replace black box models in criminal justice, healthcare and computer vision.

Key contents:

  • Key issues with explainable ML
  • Key issues with interpretable ML
  • Encouraging responsible ML governance
  • Algorithmic challenges in interpretable ML
  • Discussion on interpretability for specific domains

1/21/2021

Information Rules (資訊經營法則)



C. Shapiro and H.R. Varian, Information Rules: A Strategic Guide to the Network Economy, Harvard Business School Press, 1998.
張美惠翻譯,資訊經營法則,時報出版,2000
In Information Rules, authors Shapiro and Varian reveal that many classic economic concepts can provide the insight and understanding necessary to succeed in the information age. They argue that if managers seriously want to develop effective strategies for competing in the new economy, they must understand the fundamental economics of information technology. Whether information takes the form of software code or recorded music, is published in a book or magazine, or even posted on a website, managers must know how to evaluate the consequences of pricing, protecting, and planning new versions of information products, services, and systems. The first book to distill the economics of information and networks into practical business strategies, Information Rules is a guide to the winning moves that can help business leaders navigate successfully through the tough decisions of the information economy.
本書歸納的經濟法則歷久彌新,值得推薦給學生和專業人士了解
Chapter 1 of Information Rules begins with a description of the change brought on by technology at the close of the century--but the century described is not this one, it's the late 1800s. One hundred years ago, it was an emerging telephone and electrical network that was transforming business. Today it's the Internet. The point? While the circumstances of a particular era may be unique, the underlying principles that describe the exchange of goods in a free-market economy are the same.

1/19/2021

Dive into Deep Learning

 Aston Zhang, Zachary C. Lipton, Mu Li, and Alexander J. Smola, Dive into Deep Learning, online book.

We set out to create a resource that could (i) be freely available for everyone; (ii) offer sufficient technical depth to provide a starting point on the path to actually becoming an applied machine learning scientist; (iii) include runnable code, showing readers how to solve problems in practice; (iv) allow for rapid updates, both by us and also by the community at large; and (v) be complemented by a forum for interactive discussion of technical details and to answer questions.

1/10/2021

Interview Query for machine learning questions

Once you subscribe at their website, you will get a free weekly interview question in email. It is fun and interesting to think about these questions (as a mind game).

1/08/2021

美國華盛頓郵報的記者如何報導其老闆

Christopher Ingraham, World’s richest men added billions to their fortunes last year as others struggled, Washington Post, Jan. 1, 2021. 

America’s wealthiest, on the other hand, had a very different kind of year: Billionaires as a class have added about $1 trillion to their total net worth since the pandemic began. And roughly one-fifth of that haul flowed into the pockets of just two men: Jeff Bezos, chief executive of Amazon (and owner of The Washington Post), and Elon Musk of Tesla and SpaceX fame....

What does it mean, for instance, that two men amassed enough wealth this year to end all hunger in America (with a price tag of $25 billion, according to one estimate) eight times over? Or that the $200 billion accumulated by Bezos and Musk is greater than the amount of coronavirus relief allocated to state and local governments in the Cares Act?...

Bezos last year announced he would give $10 billion to fight climate change, and in November he announced the recipients of the first $800 million in spending on Instagram. A Washington Post analysis in June of charitable spending by the wealthiest Americans — when Bezos‘s fortune totaled $143 billion — showed he gave $100 million to Feeding America and up to $25 million for All in WA, a statewide relief effort in Washington state. For the median American, Bezos’s giving was the equivalent of donating $85 at that time.

1/07/2021

Computer Scientists Break Traveling Salesperson Record

Anna R. Karlin, Nathan Klein, Shayan Oveis Gharan, A (Slightly) Improved Approximation Algorithm for Metric TSP, arXiv:2007.01409.

Erica Klarreich, Computer Scientists Break Traveling Salesperson Record, Quanta Magazine, October 8, 2020. (Nice pictures to explain the idea)

But in 1976, Nicos Christofides came up with an algorithm that efficiently finds approximate solutions — round trips that are at most 50% longer than the best round trip....

1/06/2021

The Ride of a Lifetime (我生命中的一段歷險)

 Robert Iger, The Ride of a Lifetime: Lessons Learned from 15 Years as CEO of the Walt Disney Company, Random House, 2019.

Robert Iger became CEO of The Walt Disney Company in 2005, during a difficult time. Competition was more intense than ever and technology was changing faster than at any time in the company’s history. His vision came down to three clear ideas: Recommit to the concept that quality matters, embrace technology instead of fighting it, and think bigger—think global—and turn Disney into a stronger brand in international markets.

1/05/2021

如何完成畢業專題

根據我的經驗與觀察,寫成此文件,以幫助同學有效地完成畢業專題,學習好的模範、避免常見的錯誤。

少部份是我個人的指導方式,每位老師可能都不一樣。另外,同學展示和報告時,也納入其他老師的評語和同學的答覆,所以大部份的方法或原則適用於所有的老師。

內容包括基本觀念、建議修課、如何定題目、科技部計畫、暑假準備工作、如何讀英文 (和中文) 的論文、可能的困難與解決的方法、優良作品的特色。

如果需要列印, 請雙面列印,環保救地球  (pdf 檔)。

第一版 2008 年12 月,持續更新中。

1/04/2021

A First Course in Random Matrix Theory

Marc Potters, A First Course in Random Matrix Theory (for Physicists, Engineers and Data Scientists), Cambridge University Press, 2020.

The real world is perceived and broken down as data, models and algorithms in the eyes of physicists and engineers. Data is noisy by nature and classical statistical tools have so far been successful in dealing with relatively smaller levels of randomness. The recent emergence of Big Data and the required computing power to analyse them have rendered classical tools outdated and insufficient. Tools such as random matrix theory and the study of large sample covariance matrices can efficiently process these big data sets and help make sense of modern, deep learning algorithms. Presenting an introductory calculus course for random matrices, the book focusses on modern concepts in matrix theory, generalising the standard concept of probabilistic independence to non-commuting random variables. Concretely worked out examples and applications to financial engineering and portfolio construction make this unique book an essential tool for physicists, engineers, data analysts, and economists.

1/03/2021

Language Models are Few-Shot Learners

 Tom B. Brown et al., Language Models are Few-Shot Learners, arXiv:2005.14165.

Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.

1/02/2021

Applied Machine Learning for Embedded IoT Devices

 Vijay Janapa Reddi and Pete Warden, CS249r: Tiny Machine Learning, Harvard University, 2020 Fall.

Tiny Machine Learning (TinyML) is an introductory course at the intersection of Machine Learning and Embedded IoT Devices. The pervasiveness of ultra-low-power embedded devices, coupled with the introduction of embedded machine learning frameworks like TensorFlow Lite for Microcontrollers, will enable the mass proliferation of AI-powered IoT devices. The explosive growth in machine learning and the ease of use of platforms like TensorFlow (TF) make it an indispensable topic of study for modern computer science and electrical engineering students.