Abstract
Sensitivity properties describe how changes to the input of a program affect the output, typically by upper bounding the distance between the outputs of two runs by a monotone function of the distance between the corresponding inputs. When programs are probabilistic, the distance between outputs is a distance between distributions. The Kantorovich lifting provides a general way of defining a distance between distributions by lifting the distance of the underlying sample space; by choosing an appropriate distance on the base space, one can recover other usual probabilistic distances, such as the Total Variation distance. We develop a relational pre-expectation calculus to upper bound the Kantorovich distance between two executions of a probabilistic program. We illustrate our methods by proving algorithmic stability of a machine learning algorithm, convergence of a reinforcement learning algorithm, and fast mixing for card shuffling algorithms. We also consider some extensions: using our calculus to show convergence of Markov chains to the uniform distribution over states and an asynchronous extension to reason about pairs of program executions with different control flow.
- David Aldous. 1983. Random Walks on Finite Groups and Rapidly Mixing Markov Chains. In Séminaire de Probabilités XVII 1981/82 (Lecture Notes in Mathematics, Vol. 986 ). Springer-Verlag, 243-297. https://eudml.org/doc/113445Google Scholar
- Philip Amortila, Doina Precup, Prakash Panangaden, and Marc G. Bellemare. 2020. A Distributional Analysis of SamplingBased Reinforcement Learning Algorithms. In The 23rd International Conference on Artificial Intelligence and Statistics, AISTATS 2020, 26-28 August 2020, Online [Palermo, Sicily, Italy] (Proceedings of Machine Learning Research, Vol. 108 ), Silvia Chiappa and Roberto Calandra (Eds.). PMLR, 4357-4366. http://proceedings.mlr.press/v108/amortila20a.htmlGoogle Scholar
- Robert B. Ash and Catherine A. Doleans-Dade. 2000. Probability and Measure Theory. Academic Press.Google Scholar
- Arthur Azevedo de Amorim, Marco Gaboardi, Emilio Jesús Gallego Arias, and Justin Hsu. 2014. Really natural linear indexed type-checking. In Symposium on Implementation and Application of Functional Programming Languages (IFL), Boston, Massachusetts. ACM Press, 5 : 1-5 : 12. https://doi.org/10.1145/2746325.2746335 Google ScholarDigital Library
- Arthur Azevedo de Amorim, Marco Gaboardi, Justin Hsu, Shin-ya Katsumata, and Ikram Cherigui. 2017. A semantic account of metric preservation. In ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages (POPL), Paris, France. 545-556. https://doi.org/10.1145/3009837.3009890 Google ScholarDigital Library
- Giorgio Bacci, Radu Mardare, Prakash Panangaden, and Gordon D. Plotkin. 2018. An Algebraic Theory of Markov Processes. In Proceedings of the 33rd Annual ACM/IEEE Symposium on Logic in Computer Science, LICS 2018, Oxford, UK, July 09-12, 2018, Anuj Dawar and Erich Grädel (Eds.). ACM, 679-688. https://doi.org/10.1145/3209108.3209177 Google ScholarDigital Library
- Gilles Barthe, Thomas Espitau, Benjamin Grégoire, Justin Hsu, and Pierre-Yves Strub. 2018. Proving expected sensitivity of probabilistic programs. PACMPL 2, POPL ( 2018 ), 57 : 1-57 : 29. https://doi.org/10.1145/3158145 Google ScholarDigital Library
- Kevin Batz, Benjamin Lucien Kaminski, Joost-Pieter Katoen, Christoph Matheja, and Thomas Noll. 2019. Quantitative Separation Logic: A Logic for Reasoning About Probabilistic Pointer Programs. PACMPL 3, POPL ( 2019 ), 34 : 1-34 : 29. https://doi.org/10.1145/3290347 Google ScholarDigital Library
- Kevin Batz, Benjamin Lucien Kaminski, Joost-Pieter Katoen, and Christoph Matheja. 2021. Relatively Complete Verification of Probabilistic Programs-An Expressive Language for Expectation-based Reasoning. Proc. ACM Program. Lang. 5, POPL ( 2021 ).Google ScholarDigital Library
- Nick Benton. 2004. Simple Relational Correctness Proofs for Static Analyses and Program Transformations. In ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages (POPL), Venice, Italy. 14-25. https://doi.org/10. 1145/964001.964003 Google ScholarDigital Library
- Olivier Bousquet and André Elisseef. 2002. Stability and Generalization. Journal of Machine Learning Research 2 ( 2002 ), 499-526. http://www.jmlr.org/papers/v2/bousquet02a.htmlGoogle Scholar
- Swarat Chaudhuri, Sumit Gulwani, and Roberto Lublinerman. 2010. Continuity analysis of programs. In ACM SIGPLANSIGACT Symposium on Principles of Programming Languages (POPL), Madrid, Spain. 57-70. https://doi.org/10.1145/ 1706299.1706308 Google ScholarDigital Library
- Swarat Chaudhuri, Sumit Gulwani, and Roberto Lublinerman. 2012. Continuity and robustness of programs. Commun. ACM 55, 8 ( 2012 ), 107-115. https://doi.org/10.1145/2240236.2240262 Google ScholarDigital Library
- Marco Gaboardi, Andreas Haeberlen, Justin Hsu, Arjun Narayan, and Benjamin C. Pierce. 2013. Linear dependent types for diferential privacy. In ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages (POPL), Rome, Italy. 357-370. https://doi.org/10.1145/2429069.2429113 Google ScholarDigital Library
- Daniel Gebler, Kim G. Larsen, and Simone Tini. 2016. Compositional bisimulation metric reasoning with probabilistic process calculi. Logical Methods in Computer Science 12, 4 ( 2016 ). https://doi.org/10.2168/LMCS-12 ( 4 :12) 2016 Google ScholarCross Ref
- Daniel Gebler and Simone Tini. 2018. SOS specifications for uniformly continuous operators. J. Comput. Syst. Sci. 92 ( 2018 ), 113-151. https://doi.org/10.1016/j.jcss. 2017. 09.011 Google ScholarDigital Library
- Friedrich Gretz, Joost-Pieter Katoen, and Annabelle McIver. 2014. Operational versus weakest pre-expectation semantics for the probabilistic guarded command language. Perform. Evaluation 73 ( 2014 ), 110-132. https://doi.org/10.1016/j.peva. 2013. 11.004 Google ScholarDigital Library
- Moritz Hardt, Ben Recht, and Yoram Singer. 2016. Train faster, generalize better: Stability of stochastic gradient descent. In International Conference on Machine Learning (ICML), New York, NY (Journal of Machine Learning Research, Vol. 48 ). JMLR.org, 1225-1234. http://jmlr.org/proceedings/papers/v48/hardt16.htmlGoogle Scholar
- Zixin Huang, Zhenbang Wang, and Sasa Misailovic. 2018. PSense: Automatic Sensitivity Analysis for Probabilistic Programs. In Automated Technology for Verification and Analysis-16th International Symposium, ATVA 2018, Los Angeles, CA, USA, October 7-10, 2018, Proceedings (LNCS, Vol. 11138 ), Shuvendu K. Lahiri and Chao Wang (Eds.). Springer, 387-403. https://doi.org/10.1007/978-3-030-01090-4_23 Google ScholarCross Ref
- Benjamin Lucien Kaminski and Joost-Pieter Katoen. 2017. A weakest pre-expectation semantics for mixed-sign expectations. In LICS. IEEE Computer Society, 1-12. https://doi.org/10.1109/LICS. 2017.8005153 Google ScholarCross Ref
- Benjamin Lucien Kaminski, Joost-Pieter Katoen, Christoph Matheja, and Federico Olmedo. 2016. Weakest Precondition Reasoning for Expected Run-Times of Probabilistic Programs. In European Symposium on Programming (ESOP), Eindhoven, The Netherlands (Lecture Notes in Computer Science, Vol. 9632 ). Springer-Verlag, 364-389. https://doi.org/10.1007/978-3-662-49498-1_15 Google ScholarDigital Library
- Dexter Kozen. 1985. A Probabilistic PDL. J. Comput. System Sci. 30, 2 ( 1985 ), 162-178. https://doi.org/10.1016/ 0022-0000 ( 85 ) 90012-1 Google ScholarCross Ref
- Annabelle McIver and Carroll Morgan. 2005. Abstraction, Refinement and Proof for Probabilistic Systems. Springer.Google ScholarDigital Library
- Carroll Morgan, Annabelle McIver, and Karen Seidel. 1996. Probabilistic Predicate Transformers. ACM Transactions on Programming Languages and Systems 18, 3 ( 1996 ), 325-353.Google ScholarDigital Library
- Federico Olmedo, Friedrich Gretz, Nils Jansen, Benjamin Lucien Kaminski, Joost-Pieter Katoen, and Annabelle McIver. 2018. Conditioning in Probabilistic Programming. ACM Trans. Program. Lang. Syst. 40, 1 ( 2018 ), 4 : 1-4 : 50. https: //doi.org/10.1145/3156018 Google ScholarDigital Library
- Federico Olmedo, Benjamin Lucien Kaminski, Joost-Pieter Katoen, and Christoph Matheja. 2016. Reasoning about Recursive Probabilistic Programs. In Proceedings of the 31st Annual ACM/IEEE Symposium on Logic in Computer Science, LICS ' 16, New York, NY, USA, July 5-8, 2016, Martin Grohe, Eric Koskinen, and Natarajan Shankar (Eds.). ACM, 672-681. https://doi.org/10.1145/2933575.2935317 Google ScholarDigital Library
- David Park. 1969. Fixpoint Induction and Proofs of Program Properties. Machine Intelligence 5 ( 1969 ).Google Scholar
- Benedetto Piccoli and Francesco Rossi. 2016. On Properties of the Generalized Wasserstein Distance. Archive for Rational Mechanics and Analysis 222, 3 ( 01 Dec 2016 ), 1339-1365. https://doi.org/10.1007/s00205-016-1026-7 Google ScholarCross Ref
- Jason Reed and Benjamin C Pierce. 2010. Distance Makes the Types Grow Stronger: A Calculus for Diferential Privacy. In ACM SIGPLAN International Conference on Functional Programming (ICFP), Baltimore, Maryland. https://doi.org/10.1145/ 1863543.1863568 Google ScholarDigital Library
- Richard S. Sutton. 1988. Learning to Predict by the Methods of Temporal Diferences. Mach. Learn. 3 ( 1988 ), 9-44. https://doi.org/10.1007/BF00115009 Google ScholarCross Ref
- Cédric Villani. 2008. Optimal Transport: Old and New. Springer-Verlag.Google Scholar
- Peixin Wang, Hongfei Fu, Krishnendu Chatterjee, Yuxin Deng, and Ming Xu. 2020. Proving expected sensitivity of probabilistic programs with randomized variable-dependent termination time. Proc. ACM Program. Lang. 4, POPL ( 2020 ), 25 : 1-25 : 30. https://doi.org/10.1145/3371093 Google ScholarDigital Library
- Daniel Winograd-Cort, Andreas Haeberlen, Aaron Roth, and Benjamin C. Pierce. 2017. A framework for adaptive diferential privacy. In ACM SIGPLAN International Conference on Functional Programming (ICFP), Oxford, England. 10 : 1-10 : 29. https://doi.org/10.1145/3110254 Google ScholarDigital Library
Index Terms
- A pre-expectation calculus for probabilistic sensitivity
Recommendations
A lambda-calculus foundation for universal probabilistic programming
ICFP 2016: Proceedings of the 21st ACM SIGPLAN International Conference on Functional ProgrammingWe develop the operational semantics of an untyped probabilistic λ-calculus with continuous distributions, and both hard and soft constraints,as a foundation for universal probabilistic programming languages such as Church, Anglican, and Venture. Our ...
Bayonet: probabilistic inference for networks
PLDI 2018: Proceedings of the 39th ACM SIGPLAN Conference on Programming Language Design and ImplementationNetwork operators often need to ensure that important probabilistic properties are met, such as that the probability of network congestion is below a certain threshold. Ensuring such properties is challenging and requires both a suitable language for ...
Bayonet: probabilistic inference for networks
PLDI '18Network operators often need to ensure that important probabilistic properties are met, such as that the probability of network congestion is below a certain threshold. Ensuring such properties is challenging and requires both a suitable language for ...
Comments