Mathematical Approaches to Lower Bounds: Complexity of Proofs and Computation

Home > Workshops > 2022 > Mathematical Approaches to Lower Bounds: Complexity of Proofs and Computation

Mathematical Approaches to Lower Bounds: Complexity of Proofs and Computation

 04 - 08 Jul 2022

ICMS, Bayes Centre, Edinburgh

Scientific Organisers:

  • Olaf Beyersdorff, University of Jena
  • Rahul Santhanam, University of Oxford
  • Iddo Tzameret, Imperial College London

About:

This workshop brought together researchers working in computational and proof complexity and others working on concrete fundamental hardness questions in complexity theory from diverse perspectives, to explore, discuss and build on recently discovered results in the field. The goal was to support existing directions in the area and at the same time encourage new directions and connections. The non-exhaustive focus of the workshop was existing and new combinatorial, algebraic and probabilistic approaches, and newer logical and meta-mathematical approaches.

Computational complexity theory concerns the mathematical study of the possibilities and limitations of efficient computation. Beyond its intrinsic fundamental nature, understanding which problems are efficiently solvable has many applications in our increasingly computational world: new algorithms are important to make progress in domains such as machine learning and optimization, and new complexity lower bounds (namely, computational impossibility results) are essential for example to provably secure cryptography. Arguably the central question in computational complexity theory is the well-known P vs NP question, which asks if problems whose solutions we can efficiently verify, possess efficient algorithms to find those solutions. Equivalently, the P vs NP question asks if short proofs for mathematical theorems can be found quickly, which is a question about the nature of mathematics itself. Indeed, the centrality of this question to mathematics has been recognized by its choice as one of the seven Millennium Problems by the Clay Mathematical Institute.

Despite significant effort, progress on P and NP and related complexity lower bound problems has been slow. The goal of this workshop was to set up the basis for the emergence of new approaches to these problems as well as the combination of existing approaches in new ways. 

One theme we wished to promote was the interplay between proofs and computation in complexity theory. As an analogy, classical results establish significant links between computability theory, which studies (not necessarily efficient) solvability of computational problems, and proof theory which studies (not necessarily efficient) provability of mathematical theorems. Indeed, simple proofs of the celebrated Gödel's Incompleteness Theorems can be given by using ideas from Turing’s proof of the unsolvability of the Halting Problem. In the resource-bounded setting, where the interest is in efficient computations and proofs, similar (but more intricate) connections are believed to exist, but are relatively under-explored. Recently, there has been exciting work establishing new connections between computational complexity and proof complexity (the latter studies the question of which, predominantly propositional, statements have efficient or short proofs in a given proof system). This includes new forms of feasible interpolation that provides ways to extract computational information from short proofs; new ways to interpret proofs as certain kinds of computational objects, such as algebraic circuits, allowing the transfer of complexity lower bound techniques to lower bounds on proof size and vice versa; new `lifting' results have been found which allow obtaining lower bounds on monotone computation in a generic way from lower bounds on the well-studied Resolution proof system; connections have been found between computational pseudo-randomness and the complexity of proofs in the form of proof complexity generators; and new barriers to efficient provability of strong complexity lower bounds have been found, which can be seen as resource-bounded analogues of the classical theorems of Gödel and Chaitin.

While complexity theory is a truly interdisciplinary field which draws ideas from an increasing number of different mathematical areas, much of the research on complexity lower bounds thus far has focused on combinatorial, algebraic and probabilistic approaches - we believe that bringing in perspectives and techniques from logic can further and deepen our understanding of complexity lower bounds. Indeed, the P vs NP question is inherently meta-mathematical, and so it is reasonable to expect proof complexity-theoretic considerations to be relevant both to the truth value and provability of the P vs NP question. Moreover, we aimed to create new synergies and fresh perspectives by bringing the proof complexity and computational complexity communities together.

Some specific (non-exhaustive) themes we aimed to promote are the following:  (1) Translations between proofs and computations: exploring new forms of proof to-computation conversions, such as generalized feasible interpolation. Also extending and generalising ‘lifting’ approaches, which give equivalences between proof complexity lower bounds and computational complexity lower bounds in different settings;  (2) Algebraic and semi-algebraic proofs and computation: advancing the relations between algebraic computing, algebraic complexity and proof complexity, as well as the study of semi-algebraic proof systems with applications to approximation theory; and  (3) Logic and meta-mathematics of feasible computation: showing new formal barriers on establishing lower bounds on both proofs and computation, by exploiting the inherent meta-mathematical nature of complexity lower bound questions.

Participants: 

Olaf Beyersdorff, University of Jena, Germany
Rahul Santhanam, Oxford University
Iddo Tzameret, Imperial College London
Robert Andrews, University of Illinois at Urbana Champagne 
Paul Beame, Washington University
Arnold Beckmann, Swansea University
Levente Bodnar, University of Oxford
Benjamin Boehm, Jena University, Germany
Ilario Bonacina, UPC Universitat Politècnica de Catalunya
Sam Buss, University of Californian, San Diego 
Bruno Cavalar, University of Warwick
Anuj Dawar, Cambridge University
Susanna F. de Rezende, LTH Lund University
Yuval Filmus, Technion Haifa, Israel
Nicola Galesi Sapienza ,University Rome
Azza Gaysin, The Charles University in Prague
Mika Goos, EPFL Swisserland
Nashlen Govindasamy, Imperial College London, UK
Tuomas Hakoniemi, Imperial College London, UK
Tim Hoffmann, Jena University, Germany
Christian Ikenmeyer, University of Liverpool
Emil Jerabek, Czech Academy of Sciences
Valentine Kabanets, Simon Fraser university, Canada
Erfan Khaniki, Institute of Mathematics of the Czech Academy of Sciences
Antonina Kolokolova, Memorial University of Newfoundland
Oliver Korten, Columbia university, US
Nutan Limaye, IT University of Copenhagen
Zhenjian Lu, University of Oxford
Meena Mahajan, The Institute of Mathematical Sciences
Lauria Massimo Sapienza, Università di Roma
Rafael Mendes de Oliveira, University of Waterloo
Ian Mertz, University of Toronto
Jakob Nordstrom, University of Copenhagen and Lund University, Denmark
Joanna Ochermiak, CNRS LaBRI, Bordeaux. France
Igor Oliveira, University of Warwick
Fedor Part, Institute of Mathematics CAS
Jan Pich, University of Oxford
Pavel Pudlak, Institute of Mathematics, CAS
Alexander Razborov, University of Chicago, US
Hanlin Ren, University of Oxford
Kilian Risse, KTH Royal Institute of Technology
Robert Robere McGill, University, Canada
Amir Shpilka, Tel Aviv University
Dima Sokolov, EPFL Swisserland
Luc Spachmann Friedrich-Schiller, Universität Jena
Srikanth Srinivasan, Aahrus University, Denmark
Agnes Schleitzer, University of Jena
Neil Thapen, IM CAS
Marc Vinyals 
Luming Zhang, Imperial College London
Jiatu Li, Tsinghua University
Rahul Ilango, MIT
Marco Carmosino, IBM Research
Programme with talk abstracts

Programme:

Monday, 4 July 2022
Registration
Welcome and introduction
Proofs, Search & Algorithms
Susanna F. de Rezende , LTH Lund University Survey on automatability
Coffee break
Mika Goos, EPFL Swisserland TFNP: Collapses, separations, and characterization
Oliver Korten, Columbia university, US Connections between total search and lower bounds
Lunch break
Meena Mahajan, The Institute of Mathematical Sciences Merge Resolution: QBF proofs with inbuilt strategies
Coffee break
Christian Ikenmeyer, University of Liverpool What is in #P and what is not?
Semi-Algebraic Proofs
Ilario Bonacina, UPC Universitat Politècnica de Catalunya On the strength of semi-algebraic proof systems
Welcome reception
Tuesday, 5 July 2022
Meta-Complexity & Meta-Mathematics
Rahul Santhanam, Oxford University Survey on meta complexity
Coffee break
Igor Oliveira, University of Warwick Unprovability of circuit complexity bounds in bounded arithmetic
Hanlin Ren, University of Oxford On the range avoidance problem for circuits
Jan Pich, University of Oxford Learning algorithms versus automatability of Frege systems
Lunch break
Erfan Khaniki, Institute of Mathematics of the Czech Academy of Sciences Nisan-Wigderson generators in proof complexity: new lower bounds
Kilian Risse, KTH Royal Institute of Technology The minimum circuit size problem is hard for sum of squares
Coffee break
Panel discussion
Wednesday, 6 July 2022
Combinatorics of Proofs
Robert Robere, McGill University, Canada Survey on Lifting
Coffee break
Nicola Galesi, Sapienza University Rome Depth lower bounds in Stabbing Planes
Neil Thapen, IM CAS A simple strong tradeoff between height and size in resolution
Dmitry Sokolov, EPFL Swisserland (ONLINE) A lower bound for k-DNF Resolution on random CNFs
Lunch
Free afternoon
Thursday, 7 July 2022
Algebra, circuits, and proofs
Srikanth Srinivasan, Aahrus University, Denmark Survey on algebraic circuit complexity
Coffee break
Anuj Dawar, University of Cambridge Lower bounds for symmetric arithmetic circuits
Robert Andrews, University of Illinois at Urbana Champagne Ideals, determinants, and straightening
Tuomas Hakoniemi, Imperial College London Simple hard instances for low-depth algebraic Proofs
Lunch break
Rafael Mendes de Oliveira, University of Waterloo Succinct semi-algebraic representations and computational variants of the generalized Lax conjecture
Coffee break
Open problems session
Dinner at Blonde Restaurant
Friday, 8 July 2022
Arnold Beckmann, Swansea University Survey on bounded arithmetic
Coffee break
Pavel Pudlak, Institute of Mathematics, CAS A hierarchy of propositional proof systems
Emil Jerabek, Czech Academy of Sciences Elementary analytic functions in VTC0
Azza Gaysin, The Charles University in Prague (ONLINE) Proof complexity of CSP
Lunch break
End of workshop.