Persistent homology is a cornerstone of Topological Data Analysis (TDA), valued for its ability to track the birth and death of topological features across a sequence of spaces called a filtration. However, computing and optimizing these features in isolation can be computationally expensive and conceptually detached from other fields.
A powerful paradigm has emerged to solve this: mapping persistent homology to alternative geometric, morphological, and algebraic frameworks through exact mathematical dualities and structural equivalences. By translating topological features into dual representations—such as inverted functions, spatial complements, morphological dynamics, or discrete Morse gradients—researchers can leverage optimized algorithms and deep theoretical insights from adjacent fields.
Geometric Dualities and Function Inversion
What happens to a persistence diagram when you invert a data landscape, turning peaks into valleys? Intuitively, features that appeared early in a sublevel set filtration (valleys) now appear late in a superlevel set filtration (peaks). However, this relationship is not a simple one-to-one reflection because topological features shift dimensions depending on the geometry of the underlying space.
To formalize this, David Cohen-Steiner et al. leveraged Poincaré and Lefschetz duality to derive a precise "Symmetry Theorem." They demonstrated that negating a function on a $d$-dimensional manifold maps an $r$-dimensional topological feature to a complementary $(d-r-1)$-dimensional feature. This transformation manifests geometrically as a reflection across the minor diagonal for ordinary persistence, or through the origin for extended persistence.
This duality extends beyond function negation to spatial complements. In Alexander Duality for Functions: the Persistent Behavior of Land and Water and Shore, Michael Kerber et al. studied how the persistence diagram of a function restricted to a subset of a manifold relates to its complement (e.g., land vs. water). On a sphere, the persistence diagrams of these complementary regions are exact reflections of each other across the diagonal axis. To maintain this symmetry globally despite extreme points, the authors introduced a combinatorial coordinate-swapping operation.
To ensure these dualities hold not just for individual static diagrams but across continuous transformations of data, Maximilian Schmahl established a fully categorical framework. By introducing "lifespan functors," the author proved a natural isomorphism between absolute and relative homology, ensuring that topological dualities remain mathematically consistent when analyzing maps between different data filtrations. Similarly, Jérémy Cochoy et al. utilized vector-space duality to map persistence modules representing the disappearance of features onto those representing their appearance, unifying algebraic decomposition techniques for multi-dimensional persistence.
The Topology-Morphology Equivalence
While topologists track features via homology, image processing researchers have long used Mathematical Morphology to filter images based on "morphological dynamics"—a measure of a feature's contrast or prominence. For decades, these two fields operated independently. A series of papers has bridged this gap by proving they are mathematically identical for Morse functions.
In 1D signals, Nicolas Boutry et al. proved that pairing local extrema via path-based morphological dynamics yields the exact same pairs and values as threshold-based persistent homology. The persistence value of a topological peak is quantitatively identical to the contrast dynamics value of its paired morphological valley. The authors subsequently extended this equivalence to $n$-dimensional spaces in An Equivalence Relation Between Morphological Dynamics and Persistent Homology in n-D, showing that local minima pair with 1-saddle points, which serve as the critical "passes" where separate topological components merge.
This unification, further generalized by Laurent Najman et al., allows for the direct translation of algorithms between TDA and image processing. For instance, the classic "watershed segmentation" algorithm can be reinterpreted with strict topological guarantees. Nicolas Boutry et al. and Gilles Bertrand 0001 et al. demonstrated that the gradient vector fields of discrete Morse functions are functionally equivalent to Minimum Spanning Forests (MSFs). Consequently, computing a watershed-cut on a Morse stack is identical to finding the cut of a unique MSF, enabling linear-time topological filtering of image features based on their structural persistence.
Discrete Morse Theory and Topological Optimization
Computing persistent homology on large datasets is often bottlenecked by the size of the simplicial complex. Discrete Morse Theory (DMT) alleviates this by compressing complexes into a fraction of their size, retaining only "critical" cells. The deep synergy between these two frameworks stems from how their underlying structural pairs align.
Fabian Roll bridged this gap by showing that persistence pairs from a refined sublevel set filtration generate an algebraic gradient that completely encompasses the gradient pairs of a discrete Morse function. This structural alignment allows researchers to use DMT to accelerate persistence computations and shape reconstruction. For example, Ángel Javier Alonso et al. proved a formal topological equivalence between a novel, highly efficient Delaunay-based bifiltration and standard sublevel offset bifiltrations on point clouds, using DMT to guarantee theoretical correctness while drastically reducing computational costs.
Furthermore, this structural interplay enables optimal topological simplification. Carsten Lange et al. demonstrated that by combining DMT and persistent homology, one can eliminate noisy critical points on a surface function while staying within a strict error tolerance, achieving theoretically optimal function smoothing.
Finally, bridging these discrete structures with smooth, continuous mathematics unlocks gradient-based optimization. Jacob Leygonie et al. treated the space of persistence barcodes as a "diffeological space," establishing a unified framework for differential calculus that connects discrete simplicial filtrations to smooth Morse functions via a formal chain rule.
To make topological optimization practical, Arnur Nigmetov et al. integrated homology and cohomology to compute complete gradients for every point in a persistence diagram. While homology tracks the values where features disappear, cohomology manages their appearance. This dual integration allows optimization algorithms to take "big steps" by identifying and updating entire critical sets of simplices simultaneously, bypassing traditional black-box limitations.
Key Foundational Work
Extending Persistence Using Poincaré and Lefschetz Duality (2009) — contributed 10 concept elements to this direction.
Alexander Duality for Functions: the Persistent Behavior of Land and Water and Shore (2011) — contributed 6 concept elements to this direction.