Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2606.05245

The Extremum Stack as Optimal Memory for Rate-Independent Sequence Models: Information-Theoretic Foundations and Online Complexity

Piotr Frydrych
View original source ↗Read source document ↗Export metadata ↓

Abstract

Rate-independent sequence models respond to the ordered structure of input extrema rather than to absolute timing or token position. This principle underlies Preisach Attention, a hysteretic alternative to softmax attention in which sequence history is represented by a stack of alternating extrema generated by the classical wiping-out rule. This paper establishes the information-theoretic and online algorithmic foundations of that representation. We prove that the Preisach extremum stack is not merely a convenient implementation detail, but the complete invariant of computable rate-independent sequence functionals: a functional is rate-independent if and only if it factors through the stack. We then show that the stack is minimal in two complementary senses. In the Kolmogorov setting, the shortest exact representation answering all rate-independent queries has complexity equal to that of the stack up to an additive constant independent of sequence length and stack depth. In the Shannon setting, under any input distribution, every sufficient representation contains at least as much mutual information about the input as the stack, with equality only for representations informationally equivalent to it. Finally, we analyse the online maintenance cost of this minimal state. Although the standard stack update is amortised constant time, adversarial inputs can induce linear worst-case latency. We prove a matching output-change lower bound, show that binary search reduces boundary detection but not deletion, and give an exact finger-tree implementation with worst-case (O(\log k)) update time, where (k) is the current stack depth. The results provide a principled foundation for replacing full sequence histories or KV-cache-like memories by extremum-stack states in rate-independent neural architectures, without approximation and with bounded online latency.

Explore related subjects

cs.DSmath.FA

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Piotr Frydrych. 2026-07-13. The Extremum Stack as Optimal Memory for Rate-Independent Sequence Models: Information-Theoretic Foundations and Online Complexity. https://arxiv.org/abs/2606.05245

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Beyond Kruskal: Polynomial-Time Tensor Decomposition under the Lovitz-Petrov Condition

Identifiability criteria certify that a given tensor decomposition is a unique rank decomposition. Kruskal's classical condition is one of the best-known deterministic criteria for identifiability. However, no polynomial-time decomposition algorithm is known under the Kruskal condition, and verifying the condition itself is NP-hard. Lovitz and Petrov introduced a strictly more general identifiability condition which, in contrast, is polynomial-time verifiable, but no polynomial-time decomposition algorithm was previously known under this condition. We give a polynomial-time algorithm for tensor decomposition under the Lovitz--Petrov condition. Moreover, combining our algorithm with polynomial-time verification of the Lovitz--Petrov condition yields an efficient end-to-end certification procedure: after computing a decomposition, one can deterministically certify in polynomial time that it is unique and therefore of minimum rank. This contrasts with an arbitrary tensor decomposition, which certifies only an upper bound on the tensor rank, while determining tensor rank is NP-hard in general.

cs.DS↗
arXiv2026

Poisson Exchange Beyond Submodularity: Effective Approximation Algorithms for Offline and Online Subset Selection over Matroids

Over the past decade, a growing body of research has shown that $γ$-weak submodularity broadly arises in numerous subset selection tasks, including feature selection, neural network pruning, and video summarization. Despite its prevalence, maximizing a $γ$-weakly submodular function subject to a general matroid constraint remains challenging. To date, the only known approximation guarantee is the conservative $(1+1/γ)^{-2}$ factor established by \citet{chen2018weakly}. To improve upon this result, this paper proposes a novel algorithm called \MGPE, which repeatedly performs maximum-gain local exchanges through careful control of a non-homogeneous Poisson clock, and proves that this \MGPE\ can attain an approximation ratio arbitrarily close to $ρ_γ=1-\left(γ/(2-γ)\right)^{ \frac{γ^2}{2(1-γ)} }$. In sharp contrast to the previous guarantee, our obtained factor $ρ_γ$ not only strictly improves upon $(1+1/γ)^{-2}$ for every $γ\in(0,1]$, but also can asymptotically approach the optimal $(1-1/e)$-approximation for submodular maximization as $γ\to1$. Furthermore, we surprisingly find that when the matroid constraint reduces to a cardinality or the objective satisfies the stronger notion of $α$-weak DR-submodularity, \MGPE\ can automatically recover the tight approximation ratios of $1-e^{-γ}$ and $1-e^{-α}$, respectively. Here, $α\in(0,1]$ denotes the DR ratio.

cs.DS↗
arXiv2026

Approximating Prize-Collecting TSP below 1.556

The prize-collecting traveling salesperson problem is a variant of the metric traveling salesperson problem in which vertices may be left unvisited by paying their associated penalties. The objective is to minimize the length of the tour plus the total penalty of the unvisited vertices. Blauth, Klein, and Nägele gave the previously best-known LP-relative $1.599$-approximation. We show that a simpler version of their algorithm, obtained by omitting the splitting-off preprocessing before the tree decomposition, has an LP-relative approximation ratio of $1.555761$. The improvement comes entirely from a new analysis of the parity-correction step: a simple analysis already gives $1.56$, and the stated factor follows from a numerical parameter search with exact verification.

cs.DS↗