Discrete Morse theory

From HandWiki
Revision as of 17:51, 6 February 2024 by Jport (talk | contribs) (over-write)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Discrete Morse theory is a combinatorial adaptation of Morse theory developed by Robin Forman. The theory has various practical applications in diverse fields of applied mathematics and computer science, such as configuration spaces,[1] homology computation,[2][3] denoising,[4] mesh compression,[5] and topological data analysis.[6]

Notation regarding CW complexes

Let [math]\displaystyle{ X }[/math] be a CW complex and denote by [math]\displaystyle{ \mathcal{X} }[/math] its set of cells. Define the incidence function [math]\displaystyle{ \kappa\colon\mathcal{X} \times \mathcal{X} \to \mathbb{Z} }[/math] in the following way: given two cells [math]\displaystyle{ \sigma }[/math] and [math]\displaystyle{ \tau }[/math] in [math]\displaystyle{ \mathcal{X} }[/math], let [math]\displaystyle{ \kappa(\sigma,~\tau) }[/math] be the degree of the attaching map from the boundary of [math]\displaystyle{ \sigma }[/math] to [math]\displaystyle{ \tau }[/math]. The boundary operator is the endomorphism [math]\displaystyle{ \partial }[/math] of the free abelian group generated by [math]\displaystyle{ \mathcal{X} }[/math] defined by

[math]\displaystyle{ \partial(\sigma) = \sum_{\tau \in \mathcal{X}}\kappa(\sigma,\tau)\tau. }[/math]

It is a defining property of boundary operators that [math]\displaystyle{ \partial\circ\partial \equiv 0 }[/math]. In more axiomatic definitions[7] one can find the requirement that [math]\displaystyle{ \forall \sigma,\tau^{\prime} \in \mathcal{X} }[/math]

[math]\displaystyle{ \sum_{\tau \in \mathcal{X}} \kappa(\sigma,\tau) \kappa(\tau,\tau^{\prime}) = 0 }[/math]

which is a consequence of the above definition of the boundary operator and the requirement that [math]\displaystyle{ \partial\circ\partial \equiv 0 }[/math].

Discrete Morse functions

A real-valued function [math]\displaystyle{ \mu\colon\mathcal{X} \to \mathbb{R} }[/math] is a discrete Morse function if it satisfies the following two properties:

  1. For any cell [math]\displaystyle{ \sigma \in \mathcal{X} }[/math], the number of cells [math]\displaystyle{ \tau \in \mathcal{X} }[/math] in the boundary of [math]\displaystyle{ \sigma }[/math] which satisfy [math]\displaystyle{ \mu(\sigma) \leq \mu(\tau) }[/math] is at most one.
  2. For any cell [math]\displaystyle{ \sigma \in \mathcal{X} }[/math], the number of cells [math]\displaystyle{ \tau \in \mathcal{X} }[/math] containing [math]\displaystyle{ \sigma }[/math] in their boundary which satisfy [math]\displaystyle{ \mu(\sigma) \geq \mu(\tau) }[/math] is at most one.

It can be shown[8] that the cardinalities in the two conditions cannot both be one simultaneously for a fixed cell [math]\displaystyle{ \sigma }[/math], provided that [math]\displaystyle{ \mathcal{X} }[/math] is a regular CW complex. In this case, each cell [math]\displaystyle{ \sigma \in \mathcal{X} }[/math] can be paired with at most one exceptional cell [math]\displaystyle{ \tau \in \mathcal{X} }[/math]: either a boundary cell with larger [math]\displaystyle{ \mu }[/math] value, or a co-boundary cell with smaller [math]\displaystyle{ \mu }[/math] value. The cells which have no pairs, i.e., whose function values are strictly higher than their boundary cells and strictly lower than their co-boundary cells are called critical cells. Thus, a discrete Morse function partitions the CW complex into three distinct cell collections: [math]\displaystyle{ \mathcal{X} = \mathcal{A} \sqcup \mathcal{K} \sqcup \mathcal{Q} }[/math], where:

  1. [math]\displaystyle{ \mathcal{A} }[/math] denotes the critical cells which are unpaired,
  2. [math]\displaystyle{ \mathcal{K} }[/math] denotes cells which are paired with boundary cells, and
  3. [math]\displaystyle{ \mathcal{Q} }[/math] denotes cells which are paired with co-boundary cells.

By construction, there is a bijection of sets between [math]\displaystyle{ k }[/math]-dimensional cells in [math]\displaystyle{ \mathcal{K} }[/math] and the [math]\displaystyle{ (k-1) }[/math]-dimensional cells in [math]\displaystyle{ \mathcal{Q} }[/math], which can be denoted by [math]\displaystyle{ p^k\colon\mathcal{K}^k \to \mathcal{Q}^{k-1} }[/math] for each natural number [math]\displaystyle{ k }[/math]. It is an additional technical requirement that for each [math]\displaystyle{ K \in \mathcal{K}^k }[/math], the degree of the attaching map from the boundary of [math]\displaystyle{ K }[/math] to its paired cell [math]\displaystyle{ p^k(K) \in \mathcal{Q} }[/math] is a unit in the underlying ring of [math]\displaystyle{ \mathcal{X} }[/math]. For instance, over the integers [math]\displaystyle{ \mathbb{Z} }[/math], the only allowed values are [math]\displaystyle{ \pm 1 }[/math]. This technical requirement is guaranteed, for instance, when one assumes that [math]\displaystyle{ \mathcal{X} }[/math] is a regular CW complex over [math]\displaystyle{ \mathbb{Z} }[/math].

The fundamental result of discrete Morse theory establishes that the CW complex [math]\displaystyle{ \mathcal{X} }[/math] is isomorphic on the level of homology to a new complex [math]\displaystyle{ \mathcal{A} }[/math] consisting of only the critical cells. The paired cells in [math]\displaystyle{ \mathcal{K} }[/math] and [math]\displaystyle{ \mathcal{Q} }[/math] describe gradient paths between adjacent critical cells which can be used to obtain the boundary operator on [math]\displaystyle{ \mathcal{A} }[/math]. Some details of this construction are provided in the next section.

The Morse complex

A gradient path is a sequence of paired cells

[math]\displaystyle{ \rho = (Q_1, K_1, Q_2, K_2, \ldots, Q_M, K_M) }[/math]

satisfying [math]\displaystyle{ Q_m = p(K_m) }[/math] and [math]\displaystyle{ \kappa(K_m,~Q_{m+1}) \neq 0 }[/math]. The index of this gradient path is defined to be the integer

[math]\displaystyle{ \nu(\rho) = \frac{\prod_{m=1}^{M-1}-\kappa(K_m,Q_{m+1})}{\prod_{m=1}^{M}\kappa(K_m,Q_m)}. }[/math]

The division here makes sense because the incidence between paired cells must be [math]\displaystyle{ \pm 1 }[/math]. Note that by construction, the values of the discrete Morse function [math]\displaystyle{ \mu }[/math] must decrease across [math]\displaystyle{ \rho }[/math]. The path [math]\displaystyle{ \rho }[/math] is said to connect two critical cells [math]\displaystyle{ A,A' \in \mathcal{A} }[/math] if [math]\displaystyle{ \kappa(A,Q_1) \neq 0 \neq \kappa(K_M,A') }[/math]. This relationship may be expressed as [math]\displaystyle{ A \stackrel{\rho}{\to} A' }[/math]. The multiplicity of this connection is defined to be the integer [math]\displaystyle{ m(\rho) = \kappa(A,Q_1)\cdot\nu(\rho)\cdot\kappa(K_M,A') }[/math]. Finally, the Morse boundary operator on the critical cells [math]\displaystyle{ \mathcal{A} }[/math] is defined by

[math]\displaystyle{ \Delta(A) = \kappa(A,A') + \sum_{A \stackrel{\rho}{\to} A'}m(\rho) A' }[/math]

where the sum is taken over all gradient path connections from [math]\displaystyle{ A }[/math] to [math]\displaystyle{ A' }[/math].

Basic Results

Many of the familiar results from continuous Morse theory apply in the discrete setting.

The Morse Inequalities

Let [math]\displaystyle{ \mathcal{A} }[/math] be a Morse complex associated to the CW complex [math]\displaystyle{ \mathcal{X} }[/math]. The number [math]\displaystyle{ m_q = |\mathcal{A}_q| }[/math] of [math]\displaystyle{ q }[/math]-cells in [math]\displaystyle{ \mathcal{A} }[/math] is called the [math]\displaystyle{ q }[/math]-th Morse number. Let [math]\displaystyle{ \beta_q }[/math] denote the [math]\displaystyle{ q }[/math]-th Betti number of [math]\displaystyle{ \mathcal{X} }[/math]. Then, for any [math]\displaystyle{ N \gt 0 }[/math], the following inequalities[9] hold

[math]\displaystyle{ m_N \geq \beta_N }[/math], and
[math]\displaystyle{ m_N - m_{N-1} + \dots \pm m_0 \geq \beta_N - \beta_{N-1} + \dots \pm \beta_0 }[/math]

Moreover, the Euler characteristic [math]\displaystyle{ \chi(\mathcal{X}) }[/math] of [math]\displaystyle{ \mathcal{X} }[/math] satisfies

[math]\displaystyle{ \chi(\mathcal{X}) = m_0 - m_1 + \dots \pm m_{\dim \mathcal{X}} }[/math]

Discrete Morse Homology and Homotopy Type

Let [math]\displaystyle{ \mathcal{X} }[/math] be a regular CW complex with boundary operator [math]\displaystyle{ \partial }[/math] and a discrete Morse function [math]\displaystyle{ \mu\colon\mathcal{X} \to \mathbb{R} }[/math]. Let [math]\displaystyle{ \mathcal{A} }[/math] be the associated Morse complex with Morse boundary operator [math]\displaystyle{ \Delta }[/math]. Then, there is an isomorphism[10] of homology groups

[math]\displaystyle{ H_*(\mathcal{X},\partial) \simeq H_*(\mathcal{A},\Delta), }[/math]

and similarly for the homotopy groups.

Applications

Discrete Morse theory finds its application in molecular shape analysis,[11] skeletonization of digital images/volumes,[12] graph reconstruction from noisy data,[13] denoising noisy point clouds[14] and analysing lithic tools in archaeology.[15]

See also

References

  1. Mori, Francesca; Salvetti, Mario (2011), "(Discrete) Morse theory for Configuration spaces", Mathematical Research Letters 18 (1): 39–57, doi:10.4310/MRL.2011.v18.n1.a4, http://mrlonline.org/mrl/2011-018-001/2011-018-001-004.pdf 
  2. Perseus: the Persistent Homology software.
  3. Mischaikow, Konstantin; Nanda, Vidit (2013). "Morse Theory for Filtrations and Efficient computation of Persistent Homology". Discrete & Computational Geometry 50 (2): 330–353. doi:10.1007/s00454-013-9529-6. 
  4. Bauer, Ulrich; Lange, Carsten; Wardetzky, Max (2012). "Optimal Topological Simplification of Discrete Functions on Surfaces". Discrete & Computational Geometry 47 (2): 347–377. doi:10.1007/s00454-011-9350-z. 
  5. Lewiner, T.; Lopes, H.; Tavares, G. (2004). "Applications of Forman's discrete Morse theory to topology visualization and mesh compression". IEEE Transactions on Visualization and Computer Graphics 10 (5): 499–508. doi:10.1109/TVCG.2004.18. PMID 15794132. http://www.matmidia.mat.puc-rio.br/tomlew/pdfs/morse_apps_tvcg.pdf. 
  6. "the Topology ToolKit". https://topology-tool-kit.github.io/. 
  7. Mischaikow, Konstantin; Nanda, Vidit (2013). "Morse Theory for Filtrations and Efficient computation of Persistent Homology". Discrete & Computational Geometry 50 (2): 330–353. doi:10.1007/s00454-013-9529-6. 
  8. Forman 1998, Lemma 2.5
  9. Forman 1998, Corollaries 3.5 and 3.6
  10. Forman 1998, Theorem 7.3
  11. Cazals, F.; Chazal, F.; Lewiner, T. (2003). "Molecular shape analysis based upon the morse-smale complex and the connolly function". Proceedings of the nineteenth annual symposium on Computational geometry. ACM Press. pp. 351–360. doi:10.1145/777792.777845. ISBN 978-1-58113-663-0. http://portal.acm.org/citation.cfm?doid=777792.777845. 
  12. Delgado-Friedrichs, Olaf; Robins, Vanessa; Sheppard, Adrian (March 2015). "Skeletonization and Partitioning of Digital Images Using Discrete Morse Theory". IEEE Transactions on Pattern Analysis and Machine Intelligence 37 (3): 654–666. doi:10.1109/TPAMI.2014.2346172. ISSN 1939-3539. PMID 26353267. https://ieeexplore.ieee.org/document/6873268. 
  13. Dey, Tamal K.; Wang, Jiayuan; Wang, Yusu (2018). "Graph Reconstruction by Discrete Morse Theory". in Speckmann, Bettina; Tóth, Csaba D.. 34th International Symposium on Computational Geometry (SoCG 2018). 99. Dagstuhl, Germany: Schloss Dagstuhl–Leibniz-Zentrum fuer Informatik. pp. 31:1–31:15. doi:10.4230/LIPIcs.SoCG.2018.31. ISBN 978-3-95977-066-8. http://drops.dagstuhl.de/opus/volltexte/2018/8744. 
  14. Mukherjee, Soham (2021-09-01). "Denoising with discrete Morse theory". The Visual Computer 37 (9): 2883–94. doi:10.1007/s00371-021-02255-7. 
  15. Bullenkamp, Jan Philipp; Linsel; Mara, Hubert (2022), "Lithic Feature Identification in 3D based on Discrete Morse Theory", Proceedings of Eurographics Workshop on Graphics and Cultural Heritage (GCH) (Delft, Netherlands: Eurographics Association): 55–58, doi:10.2312/VAST/VAST10/131-138, ISBN 9783038681786, ISSN 2312-6124, https://diglib.eg.org:443/handle/10.2312/gch20221224, retrieved 2022-10-05