Write a Blog >>
ICSE 2022
Sun 8 - Fri 27 May 2022
Mon 9 May 2022 22:05 - 22:10 at ICSE room 5-even hours - Synthesis and Performance Chair(s): John Grundy
Wed 11 May 2022 13:20 - 13:25 at ICSE room 4-odd hours - Synthesis and Reverse Engineering Chair(s): Reed Milewicz

Automatic machine learning, or AutoML, holds the promise of truly democratizing the use of machine learning (ML), by substantially automating the work of data scientists. However, the huge combinatorial search space of candidate pipelines means that current AutoML techniques, generate sub-optimal pipelines, or none at all, especially on large, complex datasets. In this work we propose an AutoML technique SapientML, that can learn from a corpus of existing datasets and their human-written pipelines, and efficiently generate a high-quality pipeline for a predictive task on a new dataset. To combat the search space explosion of AutoML, SapientML employs a novel divide-and-conquer strategy realized as a three-stage program synthesis approach, that reasons on successively smaller search spaces. The first stage uses a machine-learned model to predict a set of plausible ML components to constitute a pipeline. In the second stage, this is then refined into a small pool of viable concrete pipelines using syntactic constraints derived from the corpus and the machine-learned model. Dynamically evaluating these few pipelines, in the third stage, provides the best solution. We instantiate SapientML as part of a fully automated tool-chain that creates a cleaned, labeled learning corpus by mining Kaggle, learns from it, and uses the learned models to then synthesize pipelines for new predictive tasks. We have created a training corpus of 1094 pipelines spanning 170 datasets, and evaluated SapientML on a set of 41 benchmark datasets, including 10 new, large, real-world datasets from Kaggle, and against 3 state-of-the-art AutoML tools and 2 baselines. Our evaluation shows that SapientML produces the best or comparable accuracy on 27 of the benchmarks while the second best tool fails to even produce a pipeline on 9 of the instances. This difference is amplified on the 10 most challenging benchmarks, where SapientML wins on 9 instances with the other tools failing to produce pipelines on 4 or more benchmarks.

Mon 9 May

Displayed time zone: Eastern Time (US & Canada) change

22:00 - 23:00
Synthesis and PerformanceTechnical Track / SEIP - Software Engineering in Practice at ICSE room 5-even hours
Chair(s): John Grundy Monash University
22:00
5m
Talk
Toward Among-Device AI from On-Device AI with Stream Pipelines
SEIP - Software Engineering in Practice
MyungJoo Ham Samsung Electronics, Sangjung Woo Samsung Electronics, Jaeyun Jung Samsung Electronics, Wook Song Samsung Electronics, Gichan Jang Samsung Electronics, Yongjoo Ahn Samsung Electronics, Hyoungjoo Ahn Samsung Electronics
Pre-print Media Attached
22:05
5m
Talk
SapientML: Synthesizing Machine Learning Pipelines by Learning from Human-Written Solutions
Technical Track
Ripon Saha , Akira Ura Fujitsu Ltd., Sonal Mahajan Uber Technologies Inc., Chenguang Zhu University of Texas at Austin, Linyi Li University of Illinois at Urbana-Champaign, Yang Hu The University of Texas at Austin, Hiroaki Yoshida AMD, Sarfraz Khurshid The University of Texas at Austin, Mukul Prasad Fujitsu Research of America
Pre-print Media Attached
22:10
5m
Talk
Automatic Detection of Performance Bugs in Database Systems using Equivalent Queries
Technical Track
Xinyu Liu Georgia Institute of Technology, Qi Zhou Facebook, Joy Arulraj Georgia Institute of Technology, Alessandro Orso Georgia Tech
Pre-print Media Attached

Wed 11 May

Displayed time zone: Eastern Time (US & Canada) change

13:00 - 14:00
Synthesis and Reverse EngineeringTechnical Track / Journal-First Papers at ICSE room 4-odd hours
Chair(s): Reed Milewicz Sandia National Laboratories
13:00
5m
Talk
Learning to Find Usages of Library Functions in Optimized Binaries
Journal-First Papers
Toufique Ahmed University of California at Davis, Prem Devanbu Department of Computer Science, University of California, Davis, Anand Ashok Sawant University of California, Davis
Link to publication DOI Pre-print Media Attached
13:05
5m
Talk
Dynamic Update for Synthesized GR(1) Controllers
Technical Track
Gal Amram Tel Aviv University, Shahar Maoz Tel Aviv University, Israel, Itai Segall Nokia Bell-Labs, Matan Yossef Tel Aviv University
Pre-print Media Attached
13:10
5m
Talk
Push-Button Synthesis of Watch Companions for Android Apps
Technical Track
Cong Li Nanjing University, Yanyan Jiang Nanjing University, Chang Xu Nanjing University
Link to publication DOI Pre-print Media Attached
13:15
5m
Talk
Jigsaw: Large Language Models meet Program Synthesis
Technical Track
Naman Jain Microsoft Research, Skanda Vaidyanath Stanford, Arun Iyer Microsoft Research, India, Nagarajan Natarajan Microsoft Research, India, Suresh Parthasarathy Microsoft Research, India, Sriram Rajamani Microsoft Research, Rahul Sharma Microsoft Research
Pre-print Media Attached
13:20
5m
Talk
SapientML: Synthesizing Machine Learning Pipelines by Learning from Human-Written Solutions
Technical Track
Ripon Saha , Akira Ura Fujitsu Ltd., Sonal Mahajan Uber Technologies Inc., Chenguang Zhu University of Texas at Austin, Linyi Li University of Illinois at Urbana-Champaign, Yang Hu The University of Texas at Austin, Hiroaki Yoshida AMD, Sarfraz Khurshid The University of Texas at Austin, Mukul Prasad Fujitsu Research of America
Pre-print Media Attached
13:25
5m
Talk
Static Stack-Preserving Intra-Procedural Slicing of WebAssembly BinariesBest Artifact Award
Technical Track
Quentin StiƩvenart Vrije Universiteit Brussel, David Binkley Loyola University Maryland, Coen De Roover Vrije Universiteit Brussel
DOI Pre-print Media Attached

Information for Participants
Mon 9 May 2022 22:00 - 23:00 at ICSE room 5-even hours - Synthesis and Performance Chair(s): John Grundy
Info for room ICSE room 5-even hours:

Click here to go to the room on Midspace

Wed 11 May 2022 13:00 - 14:00 at ICSE room 4-odd hours - Synthesis and Reverse Engineering Chair(s): Reed Milewicz
Info for room ICSE room 4-odd hours:

Click here to go to the room on Midspace