Abstract
We study a constraint propagation algorithm to detect infeasibility of a system of linear inequalities over continuous variables, which we call activity propagation. Each iteration of this algorithm chooses a subset of the inequalities and if it infers that some of them are always active (i.e., always hold with equality), it turns them into equalities. We show that this algorithm can be described as chaotic iterations and its fixed points can be characterized by a local consistency, in a similar way to traditional local consistency methods in CSP such as arc consistency. Via complementary slackness, activity propagation can be employed to iteratively improve a dual-feasible solution of large-scale linear programs in a primal-dual loop – a special case of this method is the Virtual Arc Consistency algorithm by Cooper et al. As our second contribution, we show that this method has the same set of fixed points as block-coordinate descent (BCD) applied to the dual linear program. While BCD is popular in large-scale optimization, its fixed points need not be global optima even for convex problems and a succinct characterization of convex problems optimally solvable by BCD remains elusive. Our result may open the way for such a characterization since it allows us to characterize BCD fixed points in terms of local consistencies.
Similar content being viewed by others
Notes
Our results also apply to general BCD methods (possibly not adhering to the relative-interior rule) if their stopping points are ‘best possible’ in the sense that no sequence of arbitrary BCD updates does not improve them – this will be made more precise later (Definition 4). However, such stopping points coincide with those that are attained by BCD with relative-interior rule.
For the top and the bottom element, we adopt notation from [24, Section 1.21] to avoid using 0 and 1 which we reserve for their numerical meaning.
That is, for any \(Q'\subseteq Q\) we have \(\bigwedge _{S}{Q'}=\bigwedge _{Q}{Q'}\). Similarly for the join.
Although the operators are actually dual closure operators according to the usual formalism, it is common to call them just ‘closure operators’ in constraint programming literature. This distinction is only technical as it can be easily corrected by considering the dual setting where the order is formally reversed.
We state chaotic iterations and related results for intensive mappings and dual closures, although they can be stated (by the duality principle) also for extensive mappings and closures.
We abbreviate the phrase ‘for every \(x\in \mathbb {R}^n\) it holds that \(Ax=b,\;x\ge 0\) implies \(x_j=0\)’ by just ‘\(Ax=b,\;x\ge 0\) implies \(x_j=0\)’, understanding that the quantifier \(\forall x\) is implicitly present.
Note the notation abuse in (8a): \((y',y^k_{[m]-B})\in \mathbb R^{[m]}\) denotes the concatenation of the components of the vectors \(y'\in \mathbb {R}^B\) and \(y^k_{[m]-B} \in \mathbb {R}^{[m]-B}\) in the right order. E.g., for \(m=5\) and \(B=\{2,3\}\), we have \(y'=(y'_2,y'_3)\), \(y^k_{[m]-B}=(y^k_1,y^k_4,y^k_5)\), and \((y',y^k_{[m]-B}) = (y^k_1,y'_2,y'_3,y^k_4,y^k_5)\).
We emphasise that this is different from the usual notion of a local minimum in optimization: here (by Definition 4), the objective in a local minimum cannot be improved by any single update () instead of an arbitrary update within some neighborhood.
E.g., turning the inequality \(x_1\le 1\) into equality means changing it to \(x_1=1\). Clearly, this can be seen as adding the inequality \(x_1\ge 1\) to the system, obtaining thus the system \(x_1\le 1\), \(x_1\ge 1\).
Recalling Definition 3, system () implies \(x_j=0\) if \(x_j=0\) holds for all x satisfying (), i.e., \(x_j=0\) is an always-active inequality in (). This can be decided, e.g., by Theorem 10 or by projecting polyhedron () onto the j-th coordinate (the projection is the singleton set \(\{0\}\) if and only if () implies \(x_j=0\)). Alternatively, one can maximize \(x_j\) subject to () and the maximum equals 0 if and only if () implies \(x_j=0\).
Recall that \({\perp }\sqsubseteq J\) for any \(J\in \mathcal {J}\), so if the elements \(J,J'\) are not comparable by \(\sqsubseteq \), they are subsets of [n], hence set union in the last case in (19) is well-defined.
Note, we cannot set \(x_2=1\) because there is no inequality \(x_2\le 1\) (or \(x_2\ge 1\)) in system () that could be made active. We say more on this in Section 5.3.
In general, it does not necessarily hold that \(p_B(\emptyset )={\perp }\). E.g., if \(b=0\) (i.e., system (16a) is homogeneous), then () is feasible even with \(J=\emptyset \) and \(p_B(\emptyset )=\emptyset \ne {\perp }\).
The primal in the pair () is of course different from the primal in the pair ().
In general and analogously to Definition 3, the inequality \(C^ix\le d_i\) is always active in the system \(Ax=b,Cx\le d\) if \(Ax=b,Cx\le d\) implies \(C^ix= d_i\).
Following on Footnote 18, this result is analogous to the fact that, for a polyhedron M, the face lattice of M is anti-isomorphic to the equality set lattice of M [50, Diagram 8.1].
References
Tseng, P. (2001). Convergence of a Block Coordinate Descent Method for Nondifferentiable Minimization. Journal Optim Theory Appl., 109(3), 475–494.
Werner T., Průša D. (2019) Relative Interior Rule in Block-Coordinate Minimization. arXiv:1910.09488 [math.OC]
Werner T., Průša D., Dlask T. (2020) Relative Interior Rule in Block-Coordinate Descent. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 7559–7567.
Wainwright, M. J., & Jordan, M. I. (2008). Graphical Models, Exponential Families, and Variational Inference. Foundations and Trends in Machine Learning, 1(1–2), 1–305.
Kappes, J. H., Andres, B., Hamprecht, F. A., Schnörr, C., Nowozin, S., Batra, D., et al. (2015). A Comparative Study of Modern Inference Techniques for Structured Discrete Energy Minimization Problems. International Journal of Computer Vision, 115(2), 155–184.
Savchynskyy, B. (2019). Discrete Graphical Models - An Optimization Perspective. Foundations and Trends in Computer Graphics and Vision, 11(3–4), 160–429.
Živný, S. (2012). The Complexity of Valued Constraint Satisfaction Problems. Cognitive Technologies: Springer.
Kovalevsky, V., & Koval, V. (1975). A diffusion algorithm for decreasing energy of max-sum labeling problem. Kiev, USSR: Glushkov Institute of Cybernetics. Unpublished.
Werner, T. (2007). A Linear Programming Approach to Max-sum Problem: A Review. IEEE Transactions on Pattern Analysis and Machine Intelligence, 29(7), 1165–1179.
Kolmogorov, V. (2006). Convergent tree-reweighted message passing for energy minimization. IEEE Transactions on Pattern Analysis and Machine Intelligence, 28(10), 1568–1583.
Globerson, A., Jaakkola, T.S. (2008) Fixing max-product: Convergent message passing algorithms for MAPLP-relaxations. In: Advances in Neural Information Processing Systems. 553–560.
Tourani, S., Shekhovtsov, A., Rother, C., Savchynskyy, B. (2018) MPLP++: Fast, parallel dual block-coordinate ascent for dense graphical models. In: Proceedings of the European Conference on Computer Vision. 251–267.
Werner, T. (2010). Revisiting the Linear Programming Relaxation Approach to Gibbs Energy Minimization and Weighted Constraint Satisfaction. IEEE Trans Pattern Analysis and Machine Intelligence., 32(8), 1474–1488.
Cooper, M. C., de Givry, S., Sanchez, M., Schiex, T., Zytnicki, M., & Werner, T. (2010). Soft arc consistency revisited. Artificial Intelligence, 174(7–8), 449–478.
Koval, V. K., & Schlesinger, M. I. (1976). Dvumernoe programmirovanie v zadachakh analiza izobrazheniy (Two-dimensional Programming in Image Analysis Problems). Automatics and Telemechanics., 8, 149–168. In Russian.
Dlask, T., Werner, T. (2020) Bounding linear programs by constraint propagation: application to Max-SAT. In: International Conference on Principles and Practice of Constraint Programming. Springer. p. 177–193.
Papadimitriou, C.H., Steiglitz, K. (1998) Combinatorial optimization: algorithms and complexity. Courier Corporation.
Dlask, T. (2020) Unit Propagation by Means of Coordinate-Wise Minimization. In: International Conference on Machine Learning, Optimization, and Data Science. Springer. 688–699.
Apt, K.R. (1997) From chaotic iteration to constraint propagation. In: International Colloquium on Automata, Languages, and Programming. Springer. 36–55.
Apt K.R. The Rough Guide to Constraint Propagation. In: Conf. on Principles and Practice of Constraint Programming. Springer; 1999. p. 1–23.
Dlask, T., Werner, T. (2020) On relation between constraint propagation and block-coordinate descent in linear programs. In: International Conference on Principles and Practice of Constraint Programming. Springer. 194–210.
Dlask, T. (2022) Block-Coordinate Descent and Local Consistencies in Linear Programming [Ph.D. thesis, available online https://dspace.cvut.cz/handle/10467/102874?locale-attribute=en]. Czech Technical University in Prague, Faculty of Electrical Engineering.
Blyth, T. S. (2005). Lattices and Ordered Algebraic Structures. London: Universitext. Springer.
Davey, B. A., & Priestley, H. A. (2002). Introduction to lattices and order. Cambridge University Press.
Nation J.B.: Notes on lattice theory. https://math.hawaii.edu/~jb/math618/Nation-LatticeTheory.pdf.
Bessiere, C. ( 2006) Constraint Propagation. In: Handbook of Constraint Programming. Elsevier.
Schrijver, A. (1998). Theory of linear and integer programming. John Wiley & Sons.
Matoušek J., Gärtner B. Understanding and using linear programming. Springer-Verlag; 2006.
Schrijver, A. (2004) Combinatorial optimization: polyhedra and efficiency. Springer Science & Business Media.
Freund, R.M., Roundy, R., Todd, M.J. (1985) Identifying the set of always-active constraints in a system of linear inequalities by a single linear program. Massachusetts Institute of Technology, Alfred P. Sloan School of Management. 1674-85.
Boyd, S., & Vandenberghe, L. (2004). Convex optimization. Cambridge University Press.
Lemaréchal, C., & Hiriart-Urruty, J. B. (2004). Fundamentals of Convex Analysis. Springer Grundlehren Text Editions: Springer Verlag, New York.
Goldman, A.J., Tucker, A.W. (1956) In: Theory of Linear Programming. Princeton University Press. 53–97.
Jansen, B., Roos, C., Terlaky, T., Vial, J.P. (1993) Interior-point methodology for linear programming: duality, sensitivity analysis and computational aspects. In: Optimization in Planning and Operation of Electric Power Systems. Springer. 57–123.
Greenberg, H. J. (1994). The use of the optimal partition in a linear programming solution for postoptimal analysis. Operations Research Letters, 15(4), 179–185.
Zhang, S. (1994) On the strictly complementary slackness relation in linear programming. In: Advances in Optimization and Approximation. Springer. 347–361.
Adler, I., & Monteiro, R. D. (1992). A geometric view of parametric linear programming. Algorithmica, 8(1), 161–176.
Mehrotra, S., & Ye, Y. (1993). Finding an interior point in the optimal face of linear programs. Mathematical Programming, 62(1), 497–515.
Telgen, J. (1983). Identifying redundant constraints and implicit equalities in systems of linear constraints. Management Science, 29(10), 1209–1222.
Greenberg, H. J. (1996). Consistency, redundancy, and implied equalities in linear systems. Annals of Mathematics and Artificial Intelligence, 17(1), 37–83.
Dlask, T., & Werner, T. (2022). Classes of linear programs solvable by coordinate-wise minimization. Annals of Mathematics and Artificial Intelligence, 90(7), 777–807.
Pisinger, D., Ropke, S. (2010) Large neighborhood search. In: Handbook of metaheuristics. Springer.
Hansen, P., Mladenović, N., Brimberg, J., Pérez, J.A.M. (2010) Variable neighborhood search. In: Handbook of metaheuristics. Springer.
Werner, T. (2017) On Coordinate Minimization of Piecewise-Affine Functions. Department of Cybernetics, Faculty of Electrical Engineering, Czech Technical University in Prague. CTU-CMP-2017-05.
Dlask, T. (2018) Minimizing Convex Piecewise-Affine Functions by Local Consistency Techniques. Master’s Thesis.
Dlask, T., & Werner, T. (2020). A Class of Linear Programs Solvable by Coordinate-Wise Minimization. In I. S. Kotsireas & P. M. Pardalos (Eds.), Learning and Intelligent Optimization (pp. 52–67). Springer.
Cohen, D. A., & Jeavons, P. G. (2017). The power of propagation: when GAC is enough. Constraints, 22(1), 3–23.
Cooper M.C., Živný S. (2016) The power of arc consistency for CSPs defined by partially-ordered forbidden patterns. In: Proceedings of the 31st Annual ACM/IEEE Symposium on Logic in Computer Science; 652–661.
Ziegler, G. M. (1994). Lectures on Polytopes. New York: Springer-Verlag.
Bachem, A., Grötschel, M. (1980) New aspects of polyhedral theory. Inst. für Ökonometrie und Operations Research.
Acknowledgements
This work has been supported by the Czech Science Foundation (grant 19-09967S), the OP VVV project CZ.02.1.01/0.0/0.0/16_019/0000765, and the Grant Agency of the Czech Technical University in Prague (grants SGS19/170/OHK3/3T/13 and SGS22/061/OHK3/1T/13).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflicts of interest
The authors have no competing interests to declare.
Appendices
Appendix A: Computing an improving feasible direction
As discussed at the beginning of Section 3.2, if () is infeasible, there exists an improving feasible direction \(\bar{y}\) satisfying (). We describe one way of obtaining such a direction based on the propagation algorithm (Algorithm 3). We remark that conditions () define a whole convex cone of improving directions and our algorithm finds one of them based on its precise implementation.
Let y be a dual-feasible point such that \(p_\mathcal {B}(\tau (y)) = {\perp }\). This implies infeasibility of (), i.e., non-optimality of y. Consider sequences \((B_l)_{l=1}^L\) and \((J_l)_{l = 1}^L\) where \(J_1 \supsetneq J_2 \supsetneq \cdots \supsetneq J_L\), \(J_1 = \tau (y)\), \(J_{l+1} = p_{B_l}(J_l)\) for every \(l \in [L-1]\), and \(p_{B_L}(J_L) = {\perp }\). To construct \(\bar{y}\), we use the primal-dual pair
and proceed as outlined in Algorithm 5. Note that the primal () is a feasibility problem identical to () if \(J_l=J\) and \(B_l=B\). Even though Algorithm 5 may seem complicated, it is easy to see that in cases when the blocks B are small, problem () is also small (and thus could even be solvable in closed-form). Correctness of Algorithm 5 is given by the following theorem.
Theorem 19
Let \(J_1=\tau (y)\). If \(p_\mathcal {B}(\tau (y))={\perp }\), Algorithm 5 returns a vector \(\bar{y}^1\) satisfying ().
Proof
We will proceed by induction, i.e., we claim that for each \(l\in [L]\), \(\bar{y}^l\) satisfies \(A_j^T\bar{y}^l \ge 0\) for all \(j\in J_{l}\) and \(b^T\bar{y}^l = b^T \bar{y}^L < 0\) is maintained during the whole algorithm. Thus, eventually \(A_j^T\bar{y}^1 \ge 0\) holds for all \(j \in J_{1} =\tau (y)\).
For the base case with \(l=L\), primal () is infeasible due to \(p_{B_L}(J_L) = {\perp }\) (see Theorem 15) and dual () is therefore unbounded since it is always feasible. Thus, there exists \(\hat{y}^L\) feasible to the dual () that satisfies \(b^T\hat{y}^L < 0\). By feasibility, \(A_j^T\hat{y}_i^L \ge 0\) for all \(j \in J_L\).
For the inductive step, let \(l \le L-1\). If condition on line 3 is not satisfied, \(A_j^T\bar{y}^l \ge 0\) holds for all \(j \in J_l\) trivially by setting \(\bar{y}^l\) equal to \(\bar{y}^{l+1}\) on line 7 due to our inductive hypothesis.
If condition on line 3 is satisfied, let us focus on (). Since \(p_{B_l}(J_l) = J_{l+1}\ne {\perp }\), primal () is feasible with optimal value 0, which is attained by all feasible solutions. Let \(x^l\) and \(\hat{y}^l\) be in the relative interior of the set of optimal solutions of the primal and dual (), respectively. Since \(x^l,\hat{y}^l\) are from the relative interior, they satisfy strict complementary slackness (see Theorem 9), i.e., \(x_j = 0 \iff A_j^T\hat{y}^l > 0\) for all \(i \in J_l\). By the last statement in Theorem 15, \(x_j = 0 \wedge A_j^T\hat{y}^l > 0\) holds for all \(j \in J_{l}-J_{l+1}\) because \(p_{B_l}(J_l) = J_{l+1}\). For completeness, \(x_j > 0 \wedge A_j^T\hat{y}^l = 0\) holds for all \(j \in J_{l+1}\).
Notice that \(\delta _l\) is well-defined because condition on line 3 was satisfied. Moreover, \(\delta _l>0\) due to both \(-A_j^T\bar{y}^{l+1}\) and \(A_j^T\hat{y}^l\) being positive by definition of \(\delta _l\).
We consider the following cases to prove that \(A_j^T\bar{y}^l \ge 0\) for all \(j\in J_l\):
-
If \(j \in J_{l+1}\), then \(A_j^T\bar{y}^{l+1} \ge 0\) by inductive hypothesis and \(A_j^T\hat{y}^l = 0\) by strict complementary slackness, hence \(A_j^T \bar{y}^l = A_j^T\bar{y}^{l+1} \ge 0\).
-
If \(j \in J_{l}-J_{l+1}\), then \(A_j^T\hat{y}^l > 0\). If \(A_j^T\bar{y}^{l+1} \ge 0\), then \(A_j^T \bar{y}^{l} = A_j^T\bar{y}^{l+1} + \delta _l A_j^T\hat{y}^l \ge 0\). On the other hand, if \(A_j^T\bar{y}^{l+1} < 0\), it holds by definition of \(\delta _l\) that \(\delta _l \ge -A_j^T\bar{y}^{l+1}/A_j^T\hat{y}^{l}\), which is after a simple reformulation equivalent to \(A_j^T \bar{y}^l = A_j^T\bar{y}^{l+1}+\delta _l A_j^T\hat{y}^l \ge 0\).
Finally, it holds by strong duality that \(b^T\bar{y}^l = 0\), which yields \(b^T\bar{y}^{l} = b^T \bar{y}^{l+1} + \delta b^T\hat{y}^l = b^T\bar{y}^{l+1} < 0\).\(\square \)
Appendix B: Faces and B-consistent sets
In this section, we explain the geometric meaning of B-consistent sets, as defined in Section 3.2.1. In detail, we will show that the set of B-consistent sets is order-isomorphic to the set of non-empty faces of the polyhedron \(X_B([n])\) (defined in (22)). Consequently, the lattice \((\mathcal {J}_B,\sqsubseteq )\) (see (18)) is isomorphic to the face lattice of \(X_B([n])\).
The faces of a convex polyhedron are usually defined using valid inequalities (or supporting hyperplanes) [49, Section 2.1]. However, faces can be also equivalently obtained by forcing subsets of inequalities to be active [29, Section 5.6]. We use this latter definition. Moreover, we define faces only for the polyhedron \(X_B([n])\) (see (22)) where \(B\subseteq [m]\) is fixed.
Definition 6
Let \(B\subseteq [m]\). A set \(F\subseteq \mathbb {R}^n\) is a face of the polyhedron \(X_B([n])\) if \(F=\emptyset \) or \(F=X_B(J)\) for some \(J\subseteq [n]\). The set of all faces of \(X_B([n])\) is denoted by
It is immediate that the set of all faces of \(X_B([n])\) is finite. Moreover, the set of all faces is closed under intersections, as shown by the following corollary.
Corollary 4
([49, Proposition 2.3]) Let \(B\subseteq [m]\). If \(F, F'\in \mathcal {F}_B\), then \(F\cap F'\in \mathcal {F}_B\).
Proof
This is clear if \(F=\emptyset \) or \(F'=\emptyset \). Otherwise, there are \(J,J'\subseteq [n]\) such that \(F=X_B(J)\) and \(F'=X_B(J')\). Hence, \(X_B(J)\cap X_B(J')=X_B(J\cap J')\in \mathcal {F}_B\).\(\square \)
Thus, the face set of the polyhedron \(X_B([n])\) forms a finite meet-semilattice w.r.t. the partial order given by set inclusion where the meet operation is set intersection. Moreover, \(X_B([n])\) is the top element of this meet-semilattice, so \((\mathcal {F}_B,\subseteq )\) is a complete lattice by Theorem 1. This is known as the face lattice [27, 49, 50].
We will now describe the connection between B-consistent sets and the faces of polyhedron \(X_B([n])\). Firstly, let us point our attention to the fact that we could require the set \(J\) to be B-consistent in Definition 6. We formulate a stronger statement in the following theorem.
Theorem 20
Let \(B\subseteq [m]\). For any non-empty \(F\in \mathcal {F}_B\), there exists a unique B-consistent set \(J\subseteq [n]\) such that \(F=X_B(J)\). Conversely, for any B-consistent set \(J\subseteq [n]\), the face \(X_B(J)\) is non-empty.
Proof
For the first part, let us show that at least one such set exists. By definition, since \(F\in \mathcal {F}_B\) is non-empty, there exists \(J'\subseteq [n]\) such that \(F=X_B(J')\). Clearly, we have that \(X_B(J')=X_B(p_B(J'))\) and \(p_B(J')\) is B-consistent by Theorem 15. To show that this set is unique, let us proceed by contradiction. Let \(J_1,J_2\subseteq [n]\) be B-consistent sets such that \(X_B(J_1)=X_B(J_2)=F\) and \(J_1\ne J_2\). Without loss of generality, assume \(J_2-J_1\ne \emptyset \) and let \(j^*\in J_2-J_1\) be arbitrary. We have that \(x_{j^*}=0\) for all \(x\in X_B(J_1)\) due to \(j^*\notin J_1\), so \(J_2\) is not B-consistent as () for \(J_2\) implies \(x_{j^*}=0\) and \(j^*\in J_2\).
For the other part, it is clear that any B-consistent set defines a face of the polyhedron by Definition 6. Moreover, this face is non-empty due to () being non-empty for any B-consistent set \(J\).\(\square \)
Following Theorem 20, \(X_B\) can be interpreted as a bijection between B-consistent sets and the set of non-empty faces of \(X_B([n])\).Footnote 18
As already noted earlier in Section 3.2.1, \(X_B\) is an isotone mapping, i.e., if \(J\subseteq J'\subseteq [n]\), then \(X_B(J)\subseteq X_B(J')\). The converse relation also holds if we restrict ourselves to B-consistent sets:
Proposition 21
Let \(B\subseteq [m]\) and \(J,J'\subseteq [n]\) be B-consistent. If \(X_B(J)\subseteq X_B(J')\), then \(J\subseteq J'\).
Proof
By contradiction: let \(X_B(J)\subseteq X_B(J')\) and \(J\nsubseteq J'\). The latter implies that there is \(j\in [n]\) such that \(j\in J\) and \(j\notin J'\). By Definition 5, there exists \(x^*\in X_B(J)\) satisfying () for J with \(x^*_j>0\). However, by definition of \(X_B(J')\) (see (22)), we have \(x_j=0\) for all \(x\in X_B(J')\) (due to \(j\notin J'\)), so \(x^*\notin X_B(J')\), which contradicts \(X_B(J)\subseteq X_B(J')\).\(\square \)
It seems natural to extend the mapping \(X_B\) to obtain the isotone bijection \(X_B'{:}\,\mathcal {J}_B \rightarrow \mathcal {F}_B\) defined by
Clearly, we have that \(J\sqsubseteq J' \iff X_B'(J)\subseteq X_B'(J')\) for any \(J,J'\in \mathcal {J}_B\). The lattices \((\mathcal {J}_B,\sqsubseteq )\) and \((\mathcal {F}_B,\subseteq )\) are therefore order-isomorphic and \(X_B'\) is a lattice isomorphism [24].Footnote 19
Remark 10
In some formalisms, \(\emptyset \) does not belong to the face lattice of some polyhedra. To be precise, if \((S,\subseteq )\) is the face lattice of some polyhedron, then \((S-\{\emptyset \},\subseteq )\) is a lattice if and only if \(\bigcap \{F\mid F\in S-\{\emptyset \}\}\ne \emptyset \), i.e., if there is a minimal non-empty face [50, Section 8]. As an example, for the non-negative orthant \(X_\emptyset ([n])=\{x\in \mathbb {R}^n\mid x\ge 0\}\), we have \(\bigcap \{F\mid F\in \mathcal {F}_\emptyset -\{\emptyset \}\}=\{0\}\) and \((\mathcal {F}_\emptyset -\{\emptyset \},\subseteq )\) is a lattice where the bottom element is \(\{0\}\) (i.e., the singleton set containing the origin).
Consequently, if \((\mathcal {F}_B-\{\emptyset \},\subseteq )\) is a lattice, then \({(\mathcal {J}_B-\{\perp \},\sqsubseteq )}\) is a lattice too. These lattices are again order-isomorphic and \(X_B\) is the lattice isomorphism.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Dlask, T., Werner, T. Activity propagation in systems of linear inequalities and its relation to block-coordinate descent in linear programs. Constraints 28, 244–276 (2023). https://doi.org/10.1007/s10601-023-09349-0
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10601-023-09349-0