iMLSE 2020

2nd International Workshop on Machine Learning Systems Engineering​

December 1, 2020, Virtual - in Conjunction with APSEC 2020
 

イベント概要

1/3

開催日

5/17 

開催​時間

14:00〜17:45

機械学習工学工学研究会
キックオフシンポジウム

我が国で唯一の機械学習とソフトウェア工学の接点である機械学習工学研究会。その最初の公式活動として、機械学習・ソフトウェア工学を代表する研究者・エンジニアが一同に会するシンポジウムを開催します。

Overview

International Workshop on Machine Learning Systems Engineering aims to bring together leading software engineers, machine learning experts and practitioners to reflect on and discuss the challenges and implications of building software for complex Artificial Intelligence (AI) systems by using Machine Learning (ML) techniques.
 
The core idea behind this workshop is a growing concern that we have as software engineers in a world where data science, deep learning, and AI are becoming increasingly pervasive. Although AI research has allowed the development of novel algorithms capable of learning new tasks, adapting to the environment, and evolving, their implementation in software systems remains challenging. From an engineering perspective, once an algorithm is implemented, it requires a solid architecture, model/data validation, proper monitoring for changes, dedicated release engineering strategies, judicious adoption of design patterns and security checks, and thorough user experience evaluation and adjustment. All these activities require a combined knowledge in software engineering, data science, and machine learning. A failure to properly address these challenges in such complex software systems can lead to catastrophic consequences. An example of such failure is the recent human toll incidence caused by the $47-million Michigan Integrated Data Automated System (MiDAS)(see https://www.bridgemi.com/public-sector/broken-human-toll-michigans-unemployment-fraud-saga) or the recent finding that simple tweaks can fool neural networks in identifying street signs (see  https://iotsecurity.eecs.umich.edu/#roadsigns ).
 
The source of emerging difficulties is the shift in the development paradigm. Classically, we have constructed software systems in a deductive way, or by writing down the rules that govern the system behaviors as program code. With machine learning techniques, we generate such rules in an inductive way from training data. This shift does not only simply require new tools that intensively deal with data but also introduces unique characteristics. The resulting system behaviors are uncertain: black-box and less explicable. They are intrinsically imperfect and it is practically impossible to reason their correctness in a deductive way.
 
Given the critical and increasing role of AI-based systems in our society, it is now imperative to engage software engineers and machine learning experts in in-depth conversations about the necessary perspectives, approaches and road-maps to address these challenges and concerns.

Workshop Date:

1st December 2020 

Venue:

Virtual

Virtual Team Meeting

Important Dates

 Paper Deadline: 25 October 2020 (Extended from 15 October 2020) 

Notification: 13 November 2020

 

Program

Keynote Speaker:

Clark Barrett (Associate Professor at Computer Science of Stanford University,  Co-Director of Center for AI Safety, USA)

Title: Towards Rigorous Verification for Safe Artificial Intelligence

Biography:

Clark Barrett joined Stanford University as an Associate Professor (Research) of Computer Science in September 2016. Before that, he was an Associate Professor of Computer Science at the Courant Institute of Mathematical Sciences at New York University.  His PhD dissertation (Stanford, 2003) introduced a novel approach to constraint solving that is now commonly known as Satisfiability Modulo Theories (SMT).  His current work focuses on the application of SMT solvers to improve the reliability and security of software, hardware, and machine learning systems.  Since 2018, Professor Barrett has served as co-director of the Stanford Center for AI Safety, whose mission includes the development of rigorous methodologies to help ensure the safety of artificial intelligence systems.  His focus in the center has primarily been on approaches and tools for formal verification of neural networks.  He has over a hundred peer-reviewed publications and has received best paper awards at DAC, ITC, FMCAD, and IJCAR.  He is an ACM Distinguished Scientist.

Invited Speaker:

Jacomo Corbo (Cofounder and chief scientist of QuantumBlack, a McKinsey company, USA)

Title: ML 2.0: New faster, cheaper paths to building machine learning systems

Biography:

Jacomo is a cofounder and the chief scientist of QuantumBlack, a McKinsey company. He helps major organizations adopt artificial intelligence at scale to radically improve their performance. As the leader of QuantumBlack’s R&D efforts in data science, Jacomo focuses on methodological innovations that can help solve clients’ most complex challenges. His work spans several industries and domain applications, from manufacturing quality on sparse machine data to price and demand forecasts to drug-safety issues to optimization and automation opportunities in how working capital is managed. Jacomo was the Chief Race Strategist for the Renault F1 Team and has led research funded by the National Science Foundation and the National Research Council of Canada, as well as several large engineering organizations. He holds a PhD in Computer Science and Masters in Applied Statistics from Harvard University, a BEng in Electrical Engineering from McGill University, and carried out postdoctoral work at the University of Pennsylvania.

Panel Discussion: Software 2.0: New paradigms for building ML & what the future holds

Chair: Foutse Khomh (Professor, Ècole Polytechnique de Montréal, Canada)

Panellists:

  • Jacomo Corbo (QuantumBlack)

  • Hiroshi Maruyama  (PFN Fellow, Preferred Networks, JAPAN)

  • Mohamed El-Geish (Director of AI, Cisco, USA)

  • Liam Paull (Assistant Professor, Université de Montréal)

Clark-barrett-small.jpg
Jacomo-Corbo-small.jpg
Detailed Program

Session 1: Keynote, Invited talks and Panel discussion
15:00-17:30 (PST) 30-November
18:00-20:30 (EST) 30-November
23:00-01:30 (UTC) 30-November / 1st December

7:00-9:30 AM (SGT) 1st December
8:00-10:30 AM (JST) 1st December

23:00-24:00(UTC) Keynote Speech: Towards Rigorous Verification for Safe Artificial Intelligence by Clark Barrett
<short break>

00:05-00:35 Invited talk: ML 2.0: New faster, cheaper paths to building machine learning systems by Jacomo Corbo
00:35-00:50 Position talks of panellists
00:50-01:30 Panel discussion: Software 2.0: New paradigms for building ML & what the future holds

Session 2: Research/Experience Talks & Poster session
5:00-6:30 AM (PST) 1st December
8:00-9:30 AM (EST) 1st December
13:00-14:30 (UTC) 1st December

21:00-22:30 (SGT) 1st December
22:00-23:30 (JST) 1st December

13:00-13:15(UTC):  Full paper

13:15-13:31:  Short paper / Tool demonstrations

13:31-13:46: Position talks

  • Machine-Learning Software-Engineering Design Patterns: Literature Review and Practitioners’ Insights (position), Hironori Washizaki, Hironori Takeuchi, Yann-Gaël Guéhéneuc, Foutse Khomh, Naotake Natori, Naohisa Shioura and Takuo Doi

  • Towards Automated DNN Repair with Minimal Regression (position), Susumu Tokumoto, Shogo Tokui and Shinji Kikuchi

  • A Study on Deep Neural Networks Repair and Repairable Data Patterns (position), Ken Matsui, Naoyasu Ubayashi, Ryosuke Sato and Yasutaka Kamei

<short break>

13:50-14:30: Poster Session at SpatialChat

 
 

SCOPE

We call for contributions that address challenges or provide practical insights in the engineering of components or systems constructed by using machine learning techniques. Topics include requirements engineering, design, construction, testing, quality assurance, operation, maintenance, and evolution, but not limited to these examples.

SUBMISSION TYPES

We call for four types of contribution, research papers, experiment reports, tool demonstrations and position talks. Each submission will be reviewed by at least three reviewers. In either case, at least one author must attend and present their work if the submission is accepted and is welcome to the poster sessions.

RESEARCH PAPERS

Research papers focus on advanced and novel theories, methodologies, or mechanisms. Submitted papers must have been neither previously accepted for publication nor concurrently submitted for review in another journal, book, conference, or workshop.

All submissions must be in English, must not exceed 10 pages, and must come in A4 paper size PDF format and conform, at time of submission, to the IEEE Conference Proceedings Formatting Guidelines (title in 24pt font and full text in 10pt font, LaTeX users must use \documentclass[10pt,conference]{IEEEtran} without including the compsoc or compsocconf option).

EXPERIENCE PAPERS

Experience papers focus on the critical challenges that the industry faces in machine learning applications, innovative solutions, and experience getting insights. Submitted papers must have been neither previously accepted for publication nor concurrently submitted for review in another journal, book, conference, or workshop.

All submissions must be in English, must not exceed 10 pages, and must come in A4 paper size PDF format and conform, at time of submission, to the IEEE Conference Proceedings Formatting Guidelines (title in 24pt font and full text in 10pt font, LaTeX users must use \documentclass[10pt,conference]{IEEEtran} without including the compsoc or compsocconf option).

SHORT RESEARCH/EXPERIENCE PAPERS

Short papers illustrate an emerging idea of innovative solutions, challenges that the industry faces in machine learning applications, or initial experience getting insights. Submitted papers must have been neither previously accepted for publication nor concurrently submitted for review in another journal, book, conference, or workshop.

All submissions must be in English, must not exceed 4 pages, and must come in A4 paper size PDF format and conform, at time of submission, to the IEEE Conference Proceedings Formatting Guidelines (title in 24pt font and full text in 10pt font, LaTeX users must use \documentclass[10pt,conference]{IEEEtran} without including the compsoc or compsocconf option).

TOOL DEMONSTRATIONS

Tool demonstrations papers provide a highly interactive venue for researchers and practitioners to demonstrate their tools and discuss them with attendees. Demonstrations should be tool-based and describe novel aspects of early prototypes or mature tools. Submitted papers must have been neither previously accepted for publication nor concurrently submitted for review in another journal, book, conference, or workshop.

All submissions must be in English, must not exceed 2 pages, and must come in A4 paper size PDF format and conform, at time of submission, to the IEEE Conference Proceedings Formatting Guidelines (title in 24pt font and full text in 10pt font, LaTeX users must use \documentclass[10pt,conference]{IEEEtran} without including the compsoc or compsocconf option).

POSITION TALKS

Position talks aim at directly speaking to the community about new and emerging ideas or practical insights and experiences. Submissions for position talks consist of the title and abstract in 500 words. The abstracts are NOT included in the workshop proceedings.

 

PROCEEDINGS

Accepted research papers, experiment reports and Tool demonstrations papers will be included in the workshop proceedings at http://ceur-ws.org/

SUBMISSION PROCEDURE

https://easychair.org/conferences/?conf=imlse2020

Call for Paper

機械学習工学キックオフシンポジウム

Organizers

Steering Committee

Program Committee