Searcharxiv⌕ Search

arXiv subjects

Franz Mayr

Publications and source records attributed to Franz Mayr.

5 recordsLinked to original sources
arXiv2024

Congruence-based Learning of Probabilistic Deterministic Finite Automata

This work studies the question of learning probabilistic deterministic automata from language models. For this purpose, it focuses on analyzing the relations defined on algebraic structures over strings by equivalences and similarities on probability distributions. We introduce a congruence that extends the classical Myhill-Nerode congruence for formal languages. This new congruence is the basis for defining regularity over language models. We present an active learning algorithm that computes the quotient with respect to this congruence whenever the language model is regular. The paper also defines the notion of recognizability for language models and shows that it coincides with regularity for congruences. For relations which are not congruences, it shows that this is not the case. Finally, it discusses the impact of this result on learning in the context of language models.

cs.FL↗
arXiv2024

Analyzing constrained LLM through PDFA-learning

We define a congruence that copes with null next-symbol probabilities that arise when the output of a language model is constrained by some means during text generation. We develop an algorithm for efficiently learning the quotient with respect to this congruence and evaluate it on case studies for analyzing statistical properties of LLM.

cs.FL↗
arXiv2022

Towards Efficient Active Learning of PDFA

We propose a new active learning algorithm for PDFA based on three main aspects: a congruence over states which takes into account next-symbol probability distributions, a quantization that copes with differences in distributions, and an efficient tree-based data structure. Experiments showed significant performance gains with respect to reference implementations.

cs.FL↗
arXiv2005

Poly-MTO, {(CH_3)_{0.92} Re O_3}_\infty, a Conducting Two-Dimensional Organometallic Oxide

Polymeric methyltrioxorhenium, {(CH_{3})_{0.92}ReO_{3}}_{\infty} (poly-MTO), is the first member of a new class of organometallic hybrids which adopts the structural pattern and physical properties of classical perovskites in two dimensions (2D). We demonstrate how the electronic structure of poly-MTO can be tailored by intercalation of organic donor molecules, such as tetrathiafulvalene (TTF) or bis-(ethylendithio)-tetrathiafulvalene (BEDT-TTF), and by the inorganic acceptor SbF$_3$. Integration of donor molecules leads to a more insulating behavior of poly-MTO, whereas SbF$_3$ insertion does not cause any significant change in the resistivity. The resistivity data of pure poly-MTO is remarkably well described by a two-dimensional electron system. Below 38 K an unusual resistivity behavior, similar to that found in doped cuprates, is observed: The resistivity initially increases approximately as $ρ\sim$ ln$(1/T$) before it changes into a $\sqrt{T}$ dependence below 2 K. As an explanation we suggest a crossover from purely two-dimensional charge-carrier diffusion within the \{ReO$_2$\}$_{\infty}$ planes at high temperatures to three-dimensional diffusion at low temperatures in a disorder-enhanced electron-electron interaction scenario (Altshuler-Aronov correction). Furthermore, a linear positive magnetoresistance was found in the insulating regime, which is caused by spatial localization of itinerant electrons at some of the Re atoms, which formally adopt a $5d^1$ electronic configuration. X-ray diffraction, IR- and ESR-studies, temperature dependent magnetization and specific heat measurements in various magnetic fields suggest that the electronic structure of poly-MTO can safely be approximated by a purely 2D conductor.

cond-mat.mtrl-sci↗