Write a Blog >>
PLDI 2021
Sun 20 - Sat 26 June 2021 Virtual Conference

This program is tentative and subject to change.

Deep Neural Networks (DNNs) have grown in popularity over the past decade and are now being used in safety-critical domains such as aircraft collision avoidance. This has motivated a large number of techniques for finding unsafe behavior in DNNs. In contrast, this paper tackles the problem of correcting a DNN once unsafe behavior is found. We introduce the provable repair problem, which is the problem of repairing a network $N$ to construct a new network $N'$ that satisfies a given specification. If the safety specification is over a finite set of points, our Provable Point Repair algorithm can find a provably minimal repair satisfying the specification, regardless of the activation functions used. For safety specifications addressing convex polytopes containing infinitely many points, our Provable Polytope Repair algorithm can find a provably minimal repair satisfying the specification for DNNs using piecewise-linear activation functions. The key insight behind both of these algorithms is the introduction of a Decoupled DNN architecture, which allows us to reduce provable repair to a linear programming problem. Our experimental results demonstrate the efficiency and effectiveness of our Provable Repair algorithms on a variety of challenging tasks.

This program is tentative and subject to change.

Conference Day
Fri 25 Jun

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

09:00 - 09:40
Talks 5A: Machine Learning and Probabilistic ProgrammingPLDI at PLDI-A +12h
09:00
5m
Talk
DeepCuts: A Deep Learning Optimization Framework for Versatile GPU Workloads
PLDI
Wookeun JungSeoul National University, Thanh Tuan DaoSeoul National University, Jaejin LeeSeoul National University
DOI
09:05
5m
Talk
Provable Repair of Deep Neural Networks
PLDI
Matthew SotoudehUniversity of California at Davis, Aditya V. ThakurUniversity of California at Davis
DOI Pre-print Media Attached
09:10
5m
Talk
DreamCoder: Bootstrapping Inductive Program Synthesis with Wake-Sleep Library Learning
PLDI
Kevin EllisCornell University, Catherine WongMassachusetts Institute of Technology, Maxwell NyeMassachusetts Institute of Technology, Mathias Sablé-MeyerPSL University; Collège de France; NeuroSpin, Lucas MoralesMassachusetts Institute of Technology, Luke HewittMassachusetts Institute of Technology, Luc CaryMassachusetts Institute of Technology, Armando Solar-LezamaMassachusetts Institute of Technology, Joshua B. TenenbaumMassachusetts Institute of Technology
DOI
09:15
5m
Talk
Specification Synthesis with Constrained Horn Clauses
PLDI
Sumanth PrabhuTCS Research, Grigory FedyukovichFlorida State University, Kumar MadhukarTCS Research, Deepak D'SouzaIISc Bangalore
DOI
09:20
5m
Talk
Compiling Stan to Generative Probabilistic Languages and Extension to Deep Probabilistic Programming
PLDI
Guillaume BaudartInria, Javier BurroniUniversity of Massachusetts Amherst, Martin HirzelIBM Research, Louis MandelIBM Research, USA, Avraham ShinnarIBM Research
DOI
09:25
5m
Talk
Sound Probabilistic Inference via Guide Types
PLDI
Di WangCarnegie Mellon University, Jan HoffmannCarnegie Mellon University, Thomas RepsUniversity of Wisconsin
DOI
09:30
5m
Talk
SPPL: Probabilistic Programming with Fast Exact Symbolic Inference
PLDI
Feras A. SaadMassachusetts Institute of Technology, Martin C. RinardMassachusetts Institute of Technology, Vikash K. MansinghkaMassachusetts Institute of Technology
DOI Pre-print
09:35
5m
Talk
Quantitative Analysis of Assertion Violations in Probabilistic Programs
PLDI
Jinyi WangShanghai Jiao Tong University, Yican SunPeking University, Hongfei FuShanghai Jiao Tong University, Krishnendu ChatterjeeIST Austria, Amir Kafshdar GoharshadyHong Kong University of Science and Technology
DOI
21:00 - 21:40
Talks 5A: Machine Learning and Probabilistic ProgrammingPLDI at PLDI-A
21:00
5m
Talk
DeepCuts: A Deep Learning Optimization Framework for Versatile GPU Workloads
PLDI
Wookeun JungSeoul National University, Thanh Tuan DaoSeoul National University, Jaejin LeeSeoul National University
DOI
21:05
5m
Talk
Provable Repair of Deep Neural Networks
PLDI
Matthew SotoudehUniversity of California at Davis, Aditya V. ThakurUniversity of California at Davis
DOI Pre-print Media Attached
21:10
5m
Talk
DreamCoder: Bootstrapping Inductive Program Synthesis with Wake-Sleep Library Learning
PLDI
Kevin EllisCornell University, Catherine WongMassachusetts Institute of Technology, Maxwell NyeMassachusetts Institute of Technology, Mathias Sablé-MeyerPSL University; Collège de France; NeuroSpin, Lucas MoralesMassachusetts Institute of Technology, Luke HewittMassachusetts Institute of Technology, Luc CaryMassachusetts Institute of Technology, Armando Solar-LezamaMassachusetts Institute of Technology, Joshua B. TenenbaumMassachusetts Institute of Technology
DOI
21:15
5m
Talk
Specification Synthesis with Constrained Horn Clauses
PLDI
Sumanth PrabhuTCS Research, Grigory FedyukovichFlorida State University, Kumar MadhukarTCS Research, Deepak D'SouzaIISc Bangalore
DOI
21:20
5m
Talk
Compiling Stan to Generative Probabilistic Languages and Extension to Deep Probabilistic Programming
PLDI
Guillaume BaudartInria, Javier BurroniUniversity of Massachusetts Amherst, Martin HirzelIBM Research, Louis MandelIBM Research, USA, Avraham ShinnarIBM Research
DOI
21:25
5m
Talk
Sound Probabilistic Inference via Guide Types
PLDI
Di WangCarnegie Mellon University, Jan HoffmannCarnegie Mellon University, Thomas RepsUniversity of Wisconsin
DOI
21:30
5m
Talk
SPPL: Probabilistic Programming with Fast Exact Symbolic Inference
PLDI
Feras A. SaadMassachusetts Institute of Technology, Martin C. RinardMassachusetts Institute of Technology, Vikash K. MansinghkaMassachusetts Institute of Technology
DOI Pre-print
21:35
5m
Talk
Quantitative Analysis of Assertion Violations in Probabilistic Programs
PLDI
Jinyi WangShanghai Jiao Tong University, Yican SunPeking University, Hongfei FuShanghai Jiao Tong University, Krishnendu ChatterjeeIST Austria, Amir Kafshdar GoharshadyHong Kong University of Science and Technology
DOI