Programming Languages
https://csd.cmu.edu/
enWed, 20 Nov 24 12:00:00 -0500Doctoral Speaking Skills Talk - Myra Dotzel
https://csd.cmu.edu/calendar/doctoral-speaking-skills-talk-myra-dotzel
<span>Doctoral Speaking Skills Talk - Myra Dotzel</span>
Gates Hillman 8102
<span><span>Anonymous (not verified)</span></span>
<span><time datetime="2024-11-20T12:00:00-05:00" title="Wednesday, November 20, 2024 - 12:00">Wed, 11/20/2024 - 12:00</time>
</span>
In Person
Modal Crash Types for Intermittent Computing
MYRA DOTZEL
<p>Intermittent computing is gaining popularity in applications that rely on batteryless energy-harvesting devices, which experience frequent and arbitrary power failures. To ensure progress, programs running on these devices rely on runtime support to save state and re-execute after a power failure. In this talk, we study the logical underpinning of intermittent computing and model checkpoint, crash, restore, and re-execution operations as computation on crash types. We draw inspiration from adjoint logic to reason about the relationship between persistent and transient memories through (re-)execution, checkpointing, and restoration. Using crash types, we show that any correct intermittent execution can be simulated by a continuously-powered execution. </p><p><em>Presented as part of the PLunch Seminar Series and in Partial Fulfillment of the CSD Speaking Skills Requirement</em></p>
https://www.andrew.cmu.edu/user/mdotzel/
Ph.D. Student, Computer Science Department, Carnegie Mellon University
https://csd.cmu.edu/calendar/doctoral-speaking-skills-talk-myra-dotzel
<a href="mailto:matthewstewart@cmu.edu">matthewstewart@cmu.edu</a>
Speaking Skills
<a href="https://csd.cmu.edu/people/doctoral-student/myra-dotzel" hreflang="en">Myra Dotzel</a>
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
Gates Hillman 8102
<p>Speaker: MYRA DOTZEL, Ph.D. Student, Computer Science Department, Carnegie Mellon University</p>
<p>Talk Title: Modal Crash Types for Intermittent Computing</p>
<p>Intermittent computing is gaining popularity in applications that rely on batteryless energy-harvesting devices, which experience frequent and arbitrary power failures. To ensure progress, programs running on these devices rely on runtime support to save state and re-execute after a power failure. In this talk, we study the logical underpinning of intermittent computing and model checkpoint, crash, restore, and re-execution operations as computation on crash types. We draw inspiration from adjoint logic to reason about the relationship between persistent and transient memories through (re-)execution, checkpointing, and restoration. Using crash types, we show that any correct intermittent execution can be simulated by a continuously-powered execution. </p>
<p>Presented as part of the PLunch Seminar Series and in Partial Fulfillment of the CSD Speaking Skills Requirement</p>
Wed, 20 Nov 2024 17:00:00 +0000Anonymous222337619 at https://csd.cmu.eduDoctoral Thesis Proposal - Long Pham
https://csd.cmu.edu/calendar/doctoral-thesis-proposal-long-pham
<span>Doctoral Thesis Proposal - Long Pham</span>
ASA Conference Room, Gates Hillman 6115
<span><span>Anonymous (not verified)</span></span>
<span><time datetime="2024-11-18T14:00:00-05:00" title="Monday, November 18, 2024 - 14:00">Mon, 11/18/2024 - 14:00</time>
</span>
In Person
Hybrid Resource-Bound Analyses of Programs
LONG PHAM
<p>Resource-bound analysis aims to infer symbolic bounds of worst-case resource usage (e.g., running time, memory, and energy) of programs as functions of program inputs. Resource analysis has numerous applications, including job scheduling in cloud computing and prevention of side-channel attacks. Various resource analysis technique have been developed, and they have unique strengths and weaknesses that complement each other. (Automatic) static resource analysis, which analyzes the source code of programs, is sound: if it successfully infers a cost bound, it is guaranteed to be a valid bound. However, every static analysis technique is incomplete: there exists a program that the analysis technique cannot handle. Meanwhile, data-driven analysis, which statistically analyzes cost measurements obtained by running programs on many inputs, can infer a candidate cost bound for any program. However, it does not guarantee soundness of inference results. </p><p>To overcome limitations of individual analysis techniques, I propose hybrid resource analysis, which integrates two complementary analysis techniques to retain their strengths while mitigating their respective weaknesses. The user first specifies which analysis techniques are used to analyze which code fragments and quantities. Hybrid analysis then performs its constituent analysis techniques on their respective code fragments and quantities. Finally, their inference results are combined into an overall cost bound. </p><p>The development of hybrid resource analysis has been driven by the desire to go beyond Automatic Amortized Resource Analysis (AARA), a state-of-the-art type-based static resource analysis technique. I start by proving polynomial-time completeness of AARA. I next introduce Bayesian data-driven analysis, which conducts Bayesian inference on cost measurements to infer a posterior distribution of symbolic cost bounds. I then present the first hybrid resource analysis, Hybrid AARA, followed by a discussion of its limitations. To overcome these limitations, I introduce the second hybrid resource analysis, resource decomposition. I additionally describe Swiftlet, which instantiates the resource-decomposition framework with AARA and Bayesian resource analysis. Finally, for proposed work, my collaborators and I plan to develop data-driven-analysis for statistically inferring not only a worst-case symbolic cost bound but also a worst-case input generator, which is a program generating worst-case program inputs of various sizes. </p><p><strong>Thesis Committee</strong></p><p>Jan Hoffmann (Chair)<br>Feras Saad<br>Matt Fredrikson<br>Nadia Polikarpova (University of California, San Diego)<br> </p><p><a href="https://www.cs.cmu.edu/~longp/publication/proposal/proposal.pdf" target="_blank">Additional Information</a></p>
<time datetime="2024-11-18T19:00:00Z">November 18, 2024 2:00pm</time>
<time datetime="2024-11-18T20:30:00Z">November 18, 2024 3:30pm</time>
https://www.cs.cmu.edu/~longp/
Ph.D. Student, Computer Science Department, Carnegie Mellon University
https://csd.cmu.edu/calendar/doctoral-thesis-proposal-long-pham
<a href="mailto:matthewstewart@cmu.edu">matthewstewart@cmu.edu</a>
Thesis Proposal
<a href="https://csd.cmu.edu/people/doctoral-student/long-pham" hreflang="en">Long Pham</a>
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
ASA Conference Room, Gates Hillman 6115
<p>Speaker: LONG PHAM, Ph.D. Student, Computer Science Department, Carnegie Mellon University</p>
<p>Talk Title: Hybrid Resource-Bound Analyses of Programs</p>
<p>Resource-bound analysis aims to infer symbolic bounds of worst-case resource usage (e.g., running time, memory, and energy) of programs as functions of program inputs. Resource analysis has numerous applications, including job scheduling in cloud computing and prevention of side-channel attacks. Various resource analysis technique have been developed, and they have unique strengths and weaknesses that complement each other. (Automatic) static resource analysis, which analyzes the source code of programs, is sound: if it successfully infers a cost bound, it is guaranteed to be a valid bound. However, every static analysis technique is incomplete: there exists a program that the analysis technique cannot handle. Meanwhile, data-driven analysis, which statistically analyzes cost measurements obtained by running programs on many inputs, can infer a candidate cost bound for any program. However, it does not guarantee soundness of inference results. </p>
<p>To overcome limitations of individual analysis techniques, I propose hybrid resource analysis, which integrates two complementary analysis techniques to retain their strengths while mitigating their respective weaknesses. The user first specifies which analysis techniques are used to analyze which code fragments and quantities. Hybrid analysis then performs its constituent analysis techniques on their respective code fragments and quantities. Finally, their inference results are combined into an overall cost bound. </p>
<p>The development of hybrid resource analysis has been driven by the desire to go beyond Automatic Amortized Resource Analysis (AARA), a state-of-the-art type-based static resource analysis technique. I start by proving polynomial-time completeness of AARA. I next introduce Bayesian data-driven analysis, which conducts Bayesian inference on cost measurements to infer a posterior distribution of symbolic cost bounds. I then present the first hybrid resource analysis, Hybrid AARA, followed by a discussion of its limitations. To overcome these limitations, I introduce the second hybrid resource analysis, resource decomposition. I additionally describe Swiftlet, which instantiates the resource-decomposition framework with AARA and Bayesian resource analysis. Finally, for proposed work, my collaborators and I plan to develop data-driven-analysis for statistically inferring not only a worst-case symbolic cost bound but also a worst-case input generator, which is a program generating worst-case program inputs of various sizes. </p>
<p>Thesis Committee</p>
<p>Jan Hoffmann (Chair)</p>
<p>Feras Saad</p>
<p>Matt Fredrikson</p>
<p>Nadia Polikarpova (University of California, San Diego)</p>
<p> </p>
<p>Additional Information</p>
Mon, 18 Nov 2024 19:00:00 +0000Anonymous222337599 at https://csd.cmu.eduDoctoral Speaking Skills Talk - Cayden Codel
https://csd.cmu.edu/calendar/doctoral-speaking-skills-talk-cayden-codel
<span>Doctoral Speaking Skills Talk - Cayden Codel</span>
Newell-Simon 3305
<span><span>Anonymous (not verified)</span></span>
<span><time datetime="2024-11-14T15:00:00-05:00" title="Thursday, November 14, 2024 - 15:00">Thu, 11/14/2024 - 15:00</time>
</span>
In Person
Verified Substitution Redundancy Checking for SAT Solving
CAYDEN CODEL
<p>One reason for the widespread adoption of SAT solvers is that they are trustworthy: their answers can be checked with verified software. In particular, many SAT solvers can emit proof certificates of unsatisfiability that are efficient to check. However, the standard proof systems in use today struggle to succinctly express proofs for problem instances with a high degree of symmetry. </p><p>In this talk, we discuss our recent work on proof checking tools for the substitution redundancy (SR) proof system. We discuss a few problems that admit short SR proofs, as well as how we can express and check those proofs. Our verified proof checker was developed in the Lean theorem prover. </p><p><em>Presented in Partial Fulfillment of the CSD Speaking Skills Requirement</em></p>
<time datetime="2024-11-14T20:00:00Z">November 14, 2024 3:00pm</time>
<time datetime="2024-11-14T21:00:00Z">November 14, 2024 4:00pm</time>
http://crcodel.com/
Ph.D. Student, Computer Science Department, Carnegie Mellon University
https://csd.cmu.edu/calendar/doctoral-speaking-skills-talk-cayden-codel
<a href="mailto:matthewstewart@cmu.edu">matthewstewart@cmu.edu</a>
Speaking Skills
<a href="https://csd.cmu.edu/people/doctoral-student/cayden-codel" hreflang="en">Cayden Codel</a>
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
Newell-Simon 3305
<p>Speaker: CAYDEN CODEL, Ph.D. Student, Computer Science Department, Carnegie Mellon University</p>
<p>Talk Title: Verified Substitution Redundancy Checking for SAT Solving</p>
<p>One reason for the widespread adoption of SAT solvers is that they are trustworthy: their answers can be checked with verified software. In particular, many SAT solvers can emit proof certificates of unsatisfiability that are efficient to check. However, the standard proof systems in use today struggle to succinctly express proofs for problem instances with a high degree of symmetry. </p>
<p>In this talk, we discuss our recent work on proof checking tools for the substitution redundancy (SR) proof system. We discuss a few problems that admit short SR proofs, as well as how we can express and check those proofs. Our verified proof checker was developed in the Lean theorem prover. </p>
<p>Presented in Partial Fulfillment of the CSD Speaking Skills Requirement</p>
Thu, 14 Nov 2024 20:00:00 +0000Anonymous222337589 at https://csd.cmu.eduDoctoral Thesis Proposal - Joseph E. Reeves
https://csd.cmu.edu/calendar/doctoral-thesis-proposal-joseph-e-reeves
<span>Doctoral Thesis Proposal - Joseph E. Reeves</span>
Reddy Conference Room, Gates Hillman 4405 and Zoom
<span><span>Anonymous (not verified)</span></span>
<span><time datetime="2024-10-29T11:00:00-04:00" title="Tuesday, October 29, 2024 - 11:00">Tue, 10/29/2024 - 11:00</time>
</span>
In Person and Virtual - ET
Cardinality Constraints in Satisfiability Solving
JOSEPH E. REEVES
<p><font>Automated reasoning (AR) engines solve problems represented in mathematics and logic stemming from a wide range of domains including hardware and software verification, cryptography, and cloud security. Boolean satisfiability (SAT) solvers drive much of the reasoning behind many AR engines, but their input format, a formula in propositional logic, can be limiting. High-level constraints must be </font><em><font>encoded</font></em><font> into sets of simpler constraints, clauses, and finding a </font><em><font>good</font></em><font> encoding often requires expert knowledge. We propose extending the input of SAT solvers to include cardinality constraints, moving encoding questions from the user-side to the solver-side. Cardinality constraints are a frequently occurring type high-level constraint that represent counting, e.g., “</font><em><font>at least k </font></em><font>packages must be delivered” or “you can work from home </font><em><font>at most one </font></em><font>day of the week”.</font></p><p><font>In this proposal we discuss four research problems arising from a cardinality-based input. First, we will develop a cardinality constraint extraction tool that will convert previously encoded problems into a cardinality-based normal form, providing backwards compatibility for our solving techniques. Second, we will engineer dynamic cardinality constraint encoding into the top-tier SAT solver CaDiCaL to improve performance on problems with many cardinality constraints. Third, we will explore the ways in which parallel solving techniques can leverage information within cardinality constraints to achieve better problem partitioning. Fourth, we will equip the extraction and solving with end-to-end proof checking through modifications to existing proof systems and proof checkers. Our goal in investigating these four research problems is to support the claim that a cardinality-based format should be the standard input format for modern SAT solvers.</font></p><p><strong>Thesis Committee</strong></p><p>Marijn Heule (Chair)<br>Randal Bryant<br>Ruben Martins<br>Armin Biere (University of Freiburg) </p><p><a href="https://www.cs.cmu.edu/~jereeves/research/proposal.pdf" target="_blank">Additional Information</a><br><br><em>In Person and </em><a href="https://cmu.zoom.us/my/jereeves" target="_blank"><em>Zoom</em></a><em> Participation. See announcement.</em></p>
<time datetime="2024-10-29T15:00:00Z">October 29, 2024 11:00am</time>
<time datetime="2024-10-29T16:30:00Z">October 29, 2024 12:30pm</time>
http://www.cs.cmu.edu/~jereeves/
Ph.D. Student, Computer Science Department, Carnegie Mellon University
https://csd.cmu.edu/calendar/doctoral-thesis-proposal-joseph-e-reeves
<a href="mailto:matthewstewart@cmu.edu">matthewstewart@cmu.edu</a>
Thesis Proposal
<a href="https://csd.cmu.edu/people/doctoral-student/joseph-reeves" hreflang="en">Joseph Reeves</a>
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
Reddy Conference Room, Gates Hillman 4405 and Zoom
<p>Speaker: JOSEPH E. REEVES, Ph.D. Student, Computer Science Department, Carnegie Mellon University</p>
<p>Talk Title: Cardinality Constraints in Satisfiability Solving</p>
<p>Automated reasoning (AR) engines solve problems represented in mathematics and logic stemming from a wide range of domains including hardware and software verification, cryptography, and cloud security. Boolean satisfiability (SAT) solvers drive much of the reasoning behind many AR engines, but their input format, a formula in propositional logic, can be limiting. High-level constraints must be encoded into sets of simpler constraints, clauses, and finding a good encoding often requires expert knowledge. We propose extending the input of SAT solvers to include cardinality constraints, moving encoding questions from the user-side to the solver-side. Cardinality constraints are a frequently occurring type high-level constraint that represent counting, e.g., “at least k packages must be delivered” or “you can work from home at most one day of the week”.</p>
<p>In this proposal we discuss four research problems arising from a cardinality-based input. First, we will develop a cardinality constraint extraction tool that will convert previously encoded problems into a cardinality-based normal form, providing backwards compatibility for our solving techniques. Second, we will engineer dynamic cardinality constraint encoding into the top-tier SAT solver CaDiCaL to improve performance on problems with many cardinality constraints. Third, we will explore the ways in which parallel solving techniques can leverage information within cardinality constraints to achieve better problem partitioning. Fourth, we will equip the extraction and solving with end-to-end proof checking through modifications to existing proof systems and proof checkers. Our goal in investigating these four research problems is to support the claim that a cardinality-based format should be the standard input format for modern SAT solvers.</p>
<p>Thesis Committee</p>
<p>Marijn Heule (Chair)</p>
<p>Randal Bryant</p>
<p>Ruben Martins</p>
<p>Armin Biere (University of Freiburg) </p>
<p>Additional Information</p>
<p>In Person and Zoom Participation. See announcement.</p>
Tue, 29 Oct 2024 15:00:00 +0000Anonymous222337267 at https://csd.cmu.eduAndre Platzer
https://csd.cmu.edu/people/faculty/andre-platzer
<span>Andre Platzer</span>
Andre
<span><span>drbarrett</span></span>
<span><time datetime="2024-10-07T17:24:04-04:00" title="Monday, October 7, 2024 - 17:24">Mon, 10/07/2024 - 17:24</time>
</span>
Platzer
Professor
Computer Science Department: Affiliated
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
<a href="https://csd.cmu.edu/people-type/faculty" hreflang="en">Faculty</a>
Off
On
Andre
Off
Mon, 07 Oct 2024 21:24:04 +0000drbarrett222337164 at https://csd.cmu.eduPrinciples of Programming Seminar - Cameron Freer
https://csd.cmu.edu/calendar/principles-of-programming-seminar-cameron-freer
<span>Principles of Programming Seminar - Cameron Freer</span>
Gates and Hillman Centers
<span><span>jennsbl</span></span>
<span><time datetime="2024-09-20T13:42:27-04:00" title="Friday, September 20, 2024 - 13:42">Fri, 09/20/2024 - 13:42</time>
</span>
6501
Computability and Symmetry in Probabilistic Programming
CAMERON FREER
<p>We consider the computable content of several key theorems in probability theory, and discuss their implications for the design of probabilistic programming languages.</p><p>A random variable is said to be exchangeable when its distribution does not depend on the ordering of the underlying elements. Sequences, arrays, graphs, and other data structures satisfying this symmetry condition are models for homogeneous data sets and serve as building blocks in Bayesian nonparametric statistics. Representation theorems by de Finetti, Aldous, Hoover, and Kallenberg show that exchangeability gives rise to conditional independence. We establish both positive and negative computable versions of these results, and explore the consequences for sequential and parallel implementations of exchangeable objects in code.</p><p>Bayes’ theorem describes conditional probabilities, and is fundamental in probabilistic inference. We show that not every computable joint distribution admits a computable conditional distribution, and examine the implications for automating Bayesian inference.</p><p>This talk is based on joint work with Nathanael Ackerman, Jeremy Avigad, Daniel Roy, and Jason Rute.</p><p>—</p><p><a href="https://cfreer.org/">Cameron Freer</a> is a Research Scientist in the Department of Brain and Cognitive Sciences at the Massachusetts Institute of Technology and a member of the MIT Probabilistic Computing Project. His research explores interactions of randomness and computation, and focuses on the foundations of probabilistic computing, efficient samplers and testing methods for probabilistic inference, and the mathematics of random structures. Cameron received his PhD in Mathematics from Harvard University advised by Gerald Sacks, and has held positions at the University of Hawaii, Keio University, and several industry research labs.</p>
<time datetime="2024-09-20T18:00:00Z">September 20, 2024 2:00pm</time>
<time datetime="2024-09-20T19:00:00Z">September 20, 2024 3:00pm</time>
https://cfreer.org/
Research Scientist, Massachusetts Institute of Technology
https://www.cs.cmu.edu/~pop/seminar/
Seminar Series
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
Gates and Hillman Centers
<p>Speaker: CAMERON FREER, Research Scientist, Massachusetts Institute of Technology</p>
<p>Talk Title: Computability and Symmetry in Probabilistic Programming</p>
<p>We consider the computable content of several key theorems in probability theory, and discuss their implications for the design of probabilistic programming languages.</p>
<p>A random variable is said to be exchangeable when its distribution does not depend on the ordering of the underlying elements. Sequences, arrays, graphs, and other data structures satisfying this symmetry condition are models for homogeneous data sets and serve as building blocks in Bayesian nonparametric statistics. Representation theorems by de Finetti, Aldous, Hoover, and Kallenberg show that exchangeability gives rise to conditional independence. We establish both positive and negative computable versions of these results, and explore the consequences for sequential and parallel implementations of exchangeable objects in code.</p>
<p>Bayes’ theorem describes conditional probabilities, and is fundamental in probabilistic inference. We show that not every computable joint distribution admits a computable conditional distribution, and examine the implications for automating Bayesian inference.</p>
<p>This talk is based on joint work with Nathanael Ackerman, Jeremy Avigad, Daniel Roy, and Jason Rute.</p>
<p>—</p>
<p>Cameron Freer is a Research Scientist in the Department of Brain and Cognitive Sciences at the Massachusetts Institute of Technology and a member of the MIT Probabilistic Computing Project. His research explores interactions of randomness and computation, and focuses on the foundations of probabilistic computing, efficient samplers and testing methods for probabilistic inference, and the mathematics of random structures. Cameron received his PhD in Mathematics from Harvard University advised by Gerald Sacks, and has held positions at the University of Hawaii, Keio University, and several industry research labs.</p>
Fri, 20 Sep 2024 17:42:27 +0000jennsbl222337049 at https://csd.cmu.eduDoctoral Thesis Proposal - Aditi Kabra
https://csd.cmu.edu/calendar/doctoral-thesis-proposal-aditi-kabra
<span>Doctoral Thesis Proposal - Aditi Kabra</span>
Gordon Bell Conference Room, Gates Hillman 5117
<span><span>Anonymous (not verified)</span></span>
<span><time datetime="2024-09-19T11:30:00-04:00" title="Thursday, September 19, 2024 - 11:30">Thu, 09/19/2024 - 11:30</time>
</span>
In Person
Verified Control Envelope Synthesis for Hybrid Systems
ADITI KABRA
<p>Many cyber-physical systems, such as trains, planes, and self-driving cars, are safety-critical but difficult to reason about. Formal verification can provide strong safety guarantees, but most industrial controllers are too complex to formally verify. <em>Safe control envelopes</em> characterize families of safe controllers and are used to monitor untrusted controllers on verifiable <em>abstractions</em> of control systems that isolate the parts relevant to safety without the full complexity of a specific control implementation, at runtime. They can put complex controllers, even when machine learning based, within the reach of formal guarantees. But correct control envelopes are still hard to design because the control engineer needs to identify correct control conditions that tell the controller what to do right now to stay safe at all times in the future by anticipating the behavior of the system over complex dynamics and an uncountably infinite state space. </p><p>This thesis proposes to provide synthesis techniques to automatically synthesize provably correct control conditions, greatly reducing the manual effort required for control envelope design. It aims to scale synthesis to complexity of real-world systems. The input of the synthesis tool is a sketch of the control envelope in a hybrid system showing what kind of control behavior is physically possible. The tool fills in the blanks of the sketch by synthesizing control conditions using hybrid system game theory. The output is a provably correct symbolic control envelope. Existing controller synthesis techniques do not solve control <em>envelope</em> synthesis because control envelopes have the higher-order constraint of permitting as many valid control solutions as possible. </p><p>Completed work provides the algorithm CESAR (Control Envelope Synthesis via Angelic Refinement) which solves a class of problems where a set of systematic game refinements allows automatic control envelope synthesis. Proposed work generalizes synthesis to a broad class of systems (characterized by admitting a natural representation in differential game logic) and develops a system that allows users to provide the human intuition based insights that, together with automated reasoning, can complete the control envelope synthesis process in more complex cases. </p><p><strong>Thesis Committee</strong></p><p>André Platzer (Co-chair, Carnegie Mellon University/Karlsruhe Institute of Technology)<br>Stefan Mitsch (Co-chair, Carnegie Mellon University/DePaul University)<br>Eunsuk Kang<br>Armando Solar-Lezama (Massachusetts Institute of Technology)<br> </p><p><a href="https://aditink.github.io/assets/proposal.pdf" target="_blank">Additional Information</a></p>
<time datetime="2024-09-19T15:30:00Z">September 19, 2024 11:30am</time>
<time datetime="2024-09-19T17:00:00Z">September 19, 2024 1:00pm</time>
https://aditink.github.io/
Ph.D. Student, Computer Science Department, Carnegie Mellon University
https://csd.cmu.edu/calendar/doctoral-thesis-proposal-aditi-kabra
<a href="mailto:matthewstewart@cmu.edu">matthewstewart@cmu.edu</a>
Thesis Proposal
<a href="https://csd.cmu.edu/people/doctoral-student/aditi-kabra" hreflang="en">Aditi Kabra</a>
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
Gordon Bell Conference Room, Gates Hillman 5117
<p>Speaker: ADITI KABRA, Ph.D. Student, Computer Science Department, Carnegie Mellon University</p>
<p>Talk Title: Verified Control Envelope Synthesis for Hybrid Systems</p>
<p>Many cyber-physical systems, such as trains, planes, and self-driving cars, are safety-critical but difficult to reason about. Formal verification can provide strong safety guarantees, but most industrial controllers are too complex to formally verify. Safe control envelopes characterize families of safe controllers and are used to monitor untrusted controllers on verifiable abstractions of control systems that isolate the parts relevant to safety without the full complexity of a specific control implementation, at runtime. They can put complex controllers, even when machine learning based, within the reach of formal guarantees. But correct control envelopes are still hard to design because the control engineer needs to identify correct control conditions that tell the controller what to do right now to stay safe at all times in the future by anticipating the behavior of the system over complex dynamics and an uncountably infinite state space. </p>
<p>This thesis proposes to provide synthesis techniques to automatically synthesize provably correct control conditions, greatly reducing the manual effort required for control envelope design. It aims to scale synthesis to complexity of real-world systems. The input of the synthesis tool is a sketch of the control envelope in a hybrid system showing what kind of control behavior is physically possible. The tool fills in the blanks of the sketch by synthesizing control conditions using hybrid system game theory. The output is a provably correct symbolic control envelope. Existing controller synthesis techniques do not solve control envelope synthesis because control envelopes have the higher-order constraint of permitting as many valid control solutions as possible. </p>
<p>Completed work provides the algorithm CESAR (Control Envelope Synthesis via Angelic Refinement) which solves a class of problems where a set of systematic game refinements allows automatic control envelope synthesis. Proposed work generalizes synthesis to a broad class of systems (characterized by admitting a natural representation in differential game logic) and develops a system that allows users to provide the human intuition based insights that, together with automated reasoning, can complete the control envelope synthesis process in more complex cases. </p>
<p>Thesis Committee</p>
<p>André Platzer (Co-chair, Carnegie Mellon University/Karlsruhe Institute of Technology)</p>
<p>Stefan Mitsch (Co-chair, Carnegie Mellon University/DePaul University)</p>
<p>Eunsuk Kang</p>
<p>Armando Solar-Lezama (Massachusetts Institute of Technology)</p>
<p> </p>
<p>Additional Information</p>
Thu, 19 Sep 2024 15:30:00 +0000Anonymous222336228 at https://csd.cmu.eduPhilosophy - Homotopy Type Theory Seminar
https://csd.cmu.edu/calendar/philosophy-homotopy-type-theory-seminar
<span>Philosophy - Homotopy Type Theory Seminar</span>
Baker Hall 150 (special date/time)
<span><span>Anonymous (not verified)</span></span>
<span><time datetime="2024-09-04T09:00:00-04:00" title="Wednesday, September 4, 2024 - 09:00">Wed, 09/04/2024 - 09:00</time>
</span>
In Person
Hofmann–Streicher lifting of fibred categories
JON STERLING
<p style="padding-left:30px;text-align:justify;"> In 1997, Hofmann and Streicher introduced an explicit technique to lift a Grothendieck universe 𝓤 from 𝐒𝐞𝐭 into the category of 𝐒𝐞𝐭-valued presheaves on a 𝓤-small category 𝓑. More recently, Awodey presented an elegant functorial analysis of this construction in terms of the ‘categorical nerve’, the right adjoint to the functor that takes a presheaf to its category of elements; in particular, applying the categorical nerve to the universal 𝓤-small discrete fibration gives the generic family of 𝓤’s Hofmann–Streicher lifting. <br> Although Awodey has investigated Hofmann–Streicher lifting in terms of a 1-functor 𝐂𝐚𝐭→𝐏𝐫(𝓑), his analysis can be extended to a 2-functor 𝐂𝐚𝐭→𝐅𝐢𝐛(𝓑) that is observed by Weber to be right 2-adjoint to the 2-functor that takes a fibred category to its total category (i.e. the oplax colimit of the corresponding diagram of categories under straightening). A generalised form of Hofmann–Streicher lifting that can be applied to categories other than universes is then obtained by conjugating this right 2-adjoint with duality involutions. <br> In joint work with Daniel Gratzer and Andrew Slattery, we have constructed a relative version of the 2-functorial Hofmann–Streicher lifting: given a fibration p:𝓐→𝓑, we have a 2-functor Δ[p]:𝐅𝐢𝐛(𝓑)→𝐅𝐢𝐛(𝓐) which is not base change but rather (we conjecture) right pseudo-adjoint to the 2-functor Σ[p]:𝐅𝐢𝐛(𝓑)→𝐅𝐢𝐛(𝓐) that sends a fibration q:𝓔→𝓐 to the composite fibration p∘q:𝓔→𝓑. A relative version of Hofmann–Streicher lifting could give a more regular theory to the practice of computing internal liftings of lifted universes.</p>
<time datetime="2024-09-04T13:00:00Z">September 4, 2024 9:00am</time>
<time datetime="2024-09-04T15:00:00Z">September 4, 2024 11:00am</time>
https://www.jonmsterling.com/
Associate Professor in Logical Foundations and Formal Methods, Department of Computer Science and Technology, University of Cambridge, and, Bye-Fellow, Clare College
https://www.cmu.edu/dietrich/philosophy/hott/seminars/index.html
Seminar Series
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
Baker Hall 150 (special date/time)
<p>Speaker: JON STERLING, Associate Professor in Logical Foundations and Formal Methods, Department of Computer Science and Technology, University of Cambridge, and, Bye-Fellow, Clare College</p>
<p>Talk Title: Hofmann–Streicher lifting of fibred categories In 1997, Hofmann and Streicher introduced an explicit technique to lift a Grothendieck universe 𝓤 from 𝐒𝐞𝐭 into the category of 𝐒𝐞𝐭-valued presheaves on a 𝓤-small category 𝓑. More recently, Awodey presented an elegant functorial analysis of this construction in terms of the ‘categorical nerve’, the right adjoint to the functor that takes a presheaf to its category of elements; in particular, applying the categorical nerve to the universal 𝓤-small discrete fibration gives the generic family of 𝓤’s Hofmann–Streicher lifting. </p>
<p> Although Awodey has investigated Hofmann–Streicher lifting in terms of a 1-functor 𝐂𝐚𝐭→𝐏𝐫(𝓑), his analysis can be extended to a 2-functor 𝐂𝐚𝐭→𝐅𝐢𝐛(𝓑) that is observed by Weber to be right 2-adjoint to the 2-functor that takes a fibred category to its total category (i.e. the oplax colimit of the corresponding diagram of categories under straightening). A generalised form of Hofmann–Streicher lifting that can be applied to categories other than universes is then obtained by conjugating this right 2-adjoint with duality involutions. </p>
<p> In joint work with Daniel Gratzer and Andrew Slattery, we have constructed a relative version of the 2-functorial Hofmann–Streicher lifting: given a fibration p:𝓐→𝓑, we have a 2-functor Δ[p]:𝐅𝐢𝐛(𝓑)→𝐅𝐢𝐛(𝓐) which is not base change but rather (we conjecture) right pseudo-adjoint to the 2-functor Σ[p]:𝐅𝐢𝐛(𝓑)→𝐅𝐢𝐛(𝓐) that sends a fibration q:𝓔→𝓐 to the composite fibration p∘q:𝓔→𝓑. A relative version of Hofmann–Streicher lifting could give a more regular theory to the practice of computing internal liftings of lifted universes.</p>
Wed, 04 Sep 2024 13:00:00 +0000Anonymous222336023 at https://csd.cmu.eduThesis Oral Defense - Jatin Arora
https://csd.cmu.edu/calendar/thesis-oral-defense-jatin-arora
<span>Thesis Oral Defense - Jatin Arora</span>
Gates Hillman 8102 and Zoom
<span><span>Anonymous (not verified)</span></span>
<span><time datetime="2024-07-29T14:00:00-04:00" title="Monday, July 29, 2024 - 14:00">Mon, 07/29/2024 - 14:00</time>
</span>
In Person and Virtual - ET
Provably Efficient Coscheduling of Computation and Data through Disentanglement
JATIN ARORA
<p>Because of its many desirable properties, such as its ability to control effects and thus potentially disastrous race conditions, functional programming offers a viable approach to programming modern multicore computers. This has led to the past decade several parallel functional languages, typically based on dialects of ML and Haskell, have been developed. These languages, however, have traditionally underperformed compared to procedural languages (such as C and Java).The primary reason for this underperformance has been the lack of scalable memory management techniques capable of matching the increased demand of memory in parallel execution.</p><p>In this thesis, we propose provably efficient techniques for memory management of parallel functional programs. The key idea behind our techniques is to coschedule the parallel computation with its data, enabling the memory manager to exploit the disentanglement hypothesis---the idea that parallel tasks of a program largely execute independently and avoid side-effecting data that may be accessed by others--for efficiency. We implement these techniques in the MPL compiler for parallel ML and our experimental results show that the techniques can marry the safety benefits of functional programming with performance.</p><p><strong>Thesis Committee:</strong></p><p>Umut A. Acar (Chair)<br>Guy E. Blelloch<br>Robert Harper<br>Rustan Leino (Amazon)<br> </p><p><em>In Person and </em><a href="https://cmu.zoom.us/j/8475808061" target="_blank"><em>Zoom</em></a><em> Participation. See announcement.</em></p>
<time datetime="2024-07-29T18:00:00Z">July 29, 2024 2:00pm</time>
<time datetime="2024-07-29T20:00:00Z">July 29, 2024 4:00pm</time>
https://www.cs.cmu.edu/~jatina/
Ph.D. Candidate, Computer Science Department, Carnegie Mellon University
https://csd.cmu.edu/calendar/thesis-oral-defense-jatin-arora
<a href="mailto:matthewstewart@cmu.edu">matthewstewart@cmu.edu</a>
Thesis Oral
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
Gates Hillman 8102 and Zoom
<p>Speaker: JATIN ARORA, Ph.D. Candidate, Computer Science Department, Carnegie Mellon University</p>
<p>Talk Title: Provably Efficient Coscheduling of Computation and Data through Disentanglement</p>
<p>Because of its many desirable properties, such as its ability to control effects and thus potentially disastrous race conditions, functional programming offers a viable approach to programming modern multicore computers. This has led to the past decade several parallel functional languages, typically based on dialects of ML and Haskell, have been developed. These languages, however, have traditionally underperformed compared to procedural languages (such as C and Java).The primary reason for this underperformance has been the lack of scalable memory management techniques capable of matching the increased demand of memory in parallel execution.</p>
<p>In this thesis, we propose provably efficient techniques for memory management of parallel functional programs. The key idea behind our techniques is to coschedule the parallel computation with its data, enabling the memory manager to exploit the disentanglement hypothesis---the idea that parallel tasks of a program largely execute independently and avoid side-effecting data that may be accessed by others--for efficiency. We implement these techniques in the MPL compiler for parallel ML and our experimental results show that the techniques can marry the safety benefits of functional programming with performance.</p>
<p>Thesis Committee:</p>
<p>Umut A. Acar (Chair)</p>
<p>Guy E. Blelloch</p>
<p>Robert Harper</p>
<p>Rustan Leino (Amazon)</p>
<p> </p>
<p>In Person and Zoom Participation. See announcement.</p>
Mon, 29 Jul 2024 18:00:00 +0000Anonymous222335614 at https://csd.cmu.eduThesis Oral Defense - Yue Niu
https://csd.cmu.edu/calendar/thesis-oral-defense-yue-niu
<span>Thesis Oral Defense - Yue Niu</span>
Gates Hillman 8102
<span><span>Anonymous (not verified)</span></span>
<span><time datetime="2024-07-29T10:00:00-04:00" title="Monday, July 29, 2024 - 10:00">Mon, 07/29/2024 - 10:00</time>
</span>
In Person
Cost-sensitive Programming, Verification, and Semantics
YUE NIU
<p>Computational cost is a fundamental aspect of the behavior of computer programs. However, existing program verification techniques do not simultaneously provide both faithful representation of cost structure and a way to reason about the pure functional meaning of cost-instrumented programs. </p><p>This thesis introduces a logical framework for integrating cost-sensitive and functional program verification and semantics by means of the internal modal type theory of presheaf categories, an approach to programming language semantics first introduced by Sterling and Harper in the context of program modules and data abstraction. I demonstrate that a range of common algorithms can be formulated and formally verified to meet both their functional and cost specifications within the framework. Lastly, I extend the logical framework and use it as a metalanguage for studying the cost semantics of programming languages, culminating in an internal cost-sensitive computational adequacy theorem for PCF that relates the denotational and operational cost semantics in the style of Plotkin. </p><p><strong>Thesis Committee:</strong> </p><p>Robert Harper (Chair)<br>Jan Hoffmann<br>Steve Brookes<br>Jon Sterling (University of Cambridge)<br>Neel Krishnaswami (University of Cambridge)</p>
<time datetime="2024-07-29T14:00:00Z">July 29, 2024 10:00am</time>
<time datetime="2024-07-29T16:00:00Z">July 29, 2024 12:00pm</time>
https://yuesforest.com/index.xml
Ph.D. Candidate, Computer Science Department, Carnegie Mellon University
https://csd.cmu.edu/calendar/thesis-oral-defense-yue-niu
<a href="mailto:matthewstewart@cmu.edu">matthewstewart@cmu.edu</a>
Thesis Oral
<a href="https://csd.cmu.edu/people/doctoral-student/yue-niu" hreflang="en">Yue Niu</a>
<a href="https://csd.cmu.edu/research/research-areas/programming-languages" hreflang="en">Programming Languages</a>
Gates HIllman 8102
<p>Speaker: YUE NIU, Ph.D. Candidate, Computer Science Department, Carnegie Mellon University</p>
<p>Talk Title: Cost-sensitive Programming, Verification, and Semantics</p>
<p>Computational cost is a fundamental aspect of the behavior of computer programs. However, existing program verification techniques do not simultaneously provide both faithful representation of cost structure and a way to reason about the pure functional meaning of cost-instrumented programs. </p>
<p>This thesis introduces a logical framework for integrating cost-sensitive and functional program verification and semantics by means of the internal modal type theory of presheaf categories, an approach to programming language semantics first introduced by Sterling and Harper in the context of program modules and data abstraction. I demonstrate that a range of common algorithms can be formulated and formally verified to meet both their functional and cost specifications within the framework. Lastly, I extend the logical framework and use it as a metalanguage for studying the cost semantics of programming languages, culminating in an internal cost-sensitive computational adequacy theorem for PCF that relates the denotational and operational cost semantics in the style of Plotkin. </p>
<p>Thesis Committee: </p>
<p>Robert Harper (Chair)</p>
<p>Jan Hoffmann</p>
<p>Steve Brookes</p>
<p>Jon Sterling (University of Cambridge)</p>
<p>Neel Krishnaswami (University of Cambridge)</p>
Mon, 29 Jul 2024 14:00:00 +0000Anonymous222335600 at https://csd.cmu.edu