Pesin [1] introduced a class of diffeomorphisms with singularities which have generalized hyperbolic attractors and established a number of properties of such diffeomorphisms. Later, Sataev [2] developed the theory of diffeomorphisms with singularities (under somewhat more restrictive conditions) up to so-called "natural" limits. However, the following problem remains open: how completely do the above-mentioned results describe the properties of nontrivial generalized hyperbolic attractors of piecewise smooth mappings that are not one-to-one? This problem is nontrivial and has been little studied yet. To the best of the author's knowledge, the paper [3] is the only one in which, among other things, the existence of a nontrivial generalized hyperbolic attractor was proved for a certain piecewise linear endomorphism of a plane. The structure of this attractor differs dramatically from that of generalized hyperbolic attractors of diffeomorphisms of the plane with singularities. The latter are one-dimensional sets with the local structure of the direct product of the Cantor set by a straight-line segment; on the opposite, the former attractor is a two-dimensional set, namely, a polygon bounded by segments of the images of the critical set of the mapping and the unstable manifold at a fixed point lying on its boundary. Therefore, by merely extending the results obtained for diffeomorphisms with singularities to the case of similar piecewise smooth endomorphisms, one cannot solve the problem posed above on the whole. The family of piecewise smooth endomorphisms of the unit square to be studied below is an example of non-oneto-one mappings whose generalized attractors have properties mainly similar to those of attractors of diffeomorphisms with singularities.Thus we consider the following family of endomorphisms of the unit square I 2 = [0, 1] ⊗ [0, 1]:Here f (t) = 2 min{t, 1 − t}, t ∈ [0, 1], T stands for transposition, and λ, 0 < λ < 1, and s > 0 are parameters. Although, in this case, the function f is only piecewise smooth, it can obviously be included in the class of so-called "unimodular mappings of the unit square" [4][5][6][7][8][9]. It is also obvious that F is the simplest representative of such mappings. 1 Before continuing the exposition of the results, we note that "unimodular mappings of the unit square" include sets of isomorphisms whose properties (for λ ∼ = 1 and for sufficiently small s > 0) are similar to those of the Henon diffeomorphism (x, y) → (1 − ax 2 + y, bx) [10] and the Lozi homeomorphism (x, y) → (1 − a|x| + y, bx) [1] (with a ∼ = 2 and sufficiently small |b| > 0). This permits one to consider the above-mentioned mappings as a possible counterpart of the Lozi and Henon mappings in the class of non-one-to-one mappings. Our main result concerning the properties of F is the existence of parameter values for which this mapping has a nontrivial one-dimensional strange attractor. From the viewpoint of dynamics, it is a generalized hyperbolic attractor, and from the viewpoint of topology, it is a...