O-minimal structures can be defined without recourse to model theory. Here we define a structure on a nonempty set M in a set-theoretic manner, as a sequence S = (Sn), n = 0,1,2,... such that
For a subset A of M, we consider the smallest structure S(A) containing S such that every finite subset of A is contained in S1. A subset D of Mn is called A-definable if it is contained in Sn(A); in that case A is called a set of parameters for D. A subset is called definable if it is A-definable for some A.
If M has a dense linear order without endpoints on it, say <, then a structure S on M is called o-minimal (respect to <) if it satisfies the extra axioms
The "o" stands for "order", since any o-minimal structure requires an ordering on the underlying set.
O-minimal structures originated in model theory and so have a simpler — but equivalent — definition using the language of model theory.2 Specifically if L is a language including a binary relation <, and (M,<,...) is an L-structure where < is interpreted to satisfy the axioms of a dense linear order,3 then (M,<,...) is called an o-minimal structure if for any definable set X ⊆ M there are finitely many open intervals I1,..., Ir in M ∪ {±∞} and a finite set X0 such that
Examples of o-minimal theories are:
In the case of RCF, the definable sets are the semialgebraic sets. Thus the study of o-minimal structures and theories generalises real algebraic geometry. A major line of current research is based on discovering expansions of the real ordered field that are o-minimal. Despite the generality of application, one can show a great deal about the geometry of set definable in o-minimal structures. There is a cell decomposition theorem,6 Whitney and Verdier stratification theorems and a good notion of dimension and Euler characteristic.
Moreover, continuously differentiable definable functions in a o-minimal structure satisfy a generalization of Łojasiewicz inequality,7 a property that has been used to guarantee the convergence of some non-smooth optimization methods, such as the stochastic subgradient method (under some mild assumptions).8910
Knight, Pillay and Steinhorn (1986), Pillay and Steinhorn (1988). ↩
Marker (2002) p.81 ↩
The condition that the interpretation of < be dense is not strictly necessary, but it is known that discrete orders lead to essentially trivial o-minimal structures, see, for example, MR0899083 and MR0943306. /wiki/MR_(identifier) ↩
Marker (2002) p.99 ↩
Patrick Speisseger, Pfaffian sets and o-minimality, in: Lecture notes on o-minimal structures and real analytic geometry, C. Miller, J.-P. Rolin, and P. Speissegger (eds.), Fields Institute Communications vol. 62, 2012, pp. 179–218. doi:10.1007/978-1-4614-4042-0_5 /wiki/Doi_(identifier) ↩
Marker (2002) p.103 ↩
Kurdyka, Krzysztof (1998). "On gradients of functions definable in o-minimal structures". Annales de l'Institut Fourier. 48 (3): 769–783. doi:10.5802/aif.1638. ISSN 0373-0956. https://aif.centre-mersenne.org/item/AIF_1998__48_3_769_0/ ↩
Davis, Damek; Drusvyatskiy, Dmitriy; Kakade, Sham; Lee, Jason D. (2020). "Stochastic Subgradient Method Converges on Tame Functions". Foundations of Computational Mathematics. 20 (1): 119–154. arXiv:1804.07795. doi:10.1007/s10208-018-09409-5. ISSN 1615-3375. S2CID 5025719. http://link.springer.com/10.1007/s10208-018-09409-5 ↩
Garrigos, Guillaume (2015-11-02). Descent dynamical systems and algorithms for tame optimization, and multi-objective problems (PhD thesis). Université Montpellier; Universidad técnica Federico Santa María (Valparaiso, Chili). https://tel.archives-ouvertes.fr/tel-02023313 ↩
Ioffe, A. D. (2009). "An Invitation to Tame Optimization". SIAM Journal on Optimization. 19 (4): 1894–1917. doi:10.1137/080722059. ISSN 1052-6234. http://epubs.siam.org/doi/10.1137/080722059 ↩