arXiv subjects

Thomas Decker

Publications and source records attributed to Thomas Decker.

At least 19 recordsLinked to original sources

arXiv2025

Improving Perturbation-based Explanations by Understanding the Role of Uncertainty Calibration

Perturbation-based explanations are widely utilized to enhance the transparency of machine-learning models in practice. However, their reliability is often compromised by the unknown model behavior under the specific perturbations used. This paper investigates the relationship between uncertainty calibration - the alignment of model confidence with actual accuracy - and perturbation-based explanations. We show that models systematically produce unreliable probability estimates when subjected to explainability-specific perturbations and theoretically prove that this directly undermines global and local explanation quality. To address this, we introduce ReCalX, a novel approach to recalibrate models for improved explanations while preserving their original predictions. Empirical evaluations across diverse models and datasets demonstrate that ReCalX consistently reduces perturbation-specific miscalibration most effectively while enhancing explanation robustness and the identification of globally important input features.

cs.LG↗

arXiv2025

Fine-Grained Uncertainty Decomposition in Large Language Models: A Spectral Approach

As Large Language Models (LLMs) are increasingly integrated in diverse applications, obtaining reliable measures of their predictive uncertainty has become critically important. A precise distinction between aleatoric uncertainty, arising from inherent ambiguities within input data, and epistemic uncertainty, originating exclusively from model limitations, is essential to effectively address each uncertainty source. In this paper, we introduce Spectral Uncertainty, a novel approach to quantifying and decomposing uncertainties in LLMs. Leveraging the Von Neumann entropy from quantum information theory, Spectral Uncertainty provides a rigorous theoretical foundation for separating total uncertainty into distinct aleatoric and epistemic components. Unlike existing baseline methods, our approach incorporates a fine-grained representation of semantic similarity, enabling nuanced differentiation among various semantic interpretations in model responses. Empirical evaluations demonstrate that Spectral Uncertainty outperforms state-of-the-art methods in estimating both aleatoric and total uncertainty across diverse models and benchmark datasets.

cs.LG↗

arXiv2025

Why Uncertainty Calibration Matters for Reliable Perturbation-based Explanations

Perturbation-based explanations are widely utilized to enhance the transparency of modern machine-learning models. However, their reliability is often compromised by the unknown model behavior under the specific perturbations used. This paper investigates the relationship between uncertainty calibration - the alignment of model confidence with actual accuracy - and perturbation-based explanations. We show that models frequently produce unreliable probability estimates when subjected to explainability-specific perturbations and theoretically prove that this directly undermines explanation quality. To address this, we introduce ReCalX, a novel approach to recalibrate models for improved perturbation-based explanations while preserving their original predictions. Experiments on popular computer vision models demonstrate that our calibration strategy produces explanations that are more aligned with human perception and actual object locations.

cs.LG↗

arXiv2025

Incremental Uncertainty-aware Performance Monitoring with Active Labeling Intervention

We study the problem of monitoring machine learning models under gradual distribution shifts, where circumstances change slowly over time, often leading to unnoticed yet significant declines in accuracy. To address this, we propose Incremental Uncertainty-aware Performance Monitoring (IUPM), a novel label-free method that estimates performance changes by modeling gradual shifts using optimal transport. In addition, IUPM quantifies the uncertainty in the performance prediction and introduces an active labeling procedure to restore a reliable estimate under a limited labeling budget. Our experiments show that IUPM outperforms existing performance estimation baselines in various gradual shift scenarios and that its uncertainty awareness guides label acquisition more effectively compared to other strategies.

cs.LG↗

arXiv2024

Explanatory Model Monitoring to Understand the Effects of Feature Shifts on Performance

Monitoring and maintaining machine learning models are among the most critical challenges in translating recent advances in the field into real-world applications. However, current monitoring methods lack the capability of provide actionable insights answering the question of why the performance of a particular model really degraded. In this work, we propose a novel approach to explain the behavior of a black-box model under feature shifts by attributing an estimated performance change to interpretable input characteristics. We refer to our method that combines concepts from Optimal Transport and Shapley Values as Explanatory Performance Estimation (XPE). We analyze the underlying assumptions and demonstrate the superiority of our approach over several baselines on different data sets across various data modalities such as images, audio, and tabular data. We also indicate how the generated results can lead to valuable insights, enabling explanatory model monitoring by revealing potential root causes for model deterioration and guiding toward actionable countermeasures.

cs.LG↗

arXiv2024

Provably Better Explanations with Optimized Aggregation of Feature Attributions

Using feature attributions for post-hoc explanations is a common practice to understand and verify the predictions of opaque machine learning models. Despite the numerous techniques available, individual methods often produce inconsistent and unstable results, putting their overall reliability into question. In this work, we aim to systematically improve the quality of feature attributions by combining multiple explanations across distinct methods or their variations. For this purpose, we propose a novel approach to derive optimal convex combinations of feature attributions that yield provable improvements of desired quality criteria such as robustness or faithfulness to the model behavior. Through extensive experiments involving various model architectures and popular feature attribution techniques, we demonstrate that our combination strategy consistently outperforms individual methods and existing baselines.

cs.LG↗

arXiv2023

Does Your Model Think Like an Engineer? Explainable AI for Bearing Fault Detection with Deep Learning

Deep Learning has already been successfully applied to analyze industrial sensor data in a variety of relevant use cases. However, the opaque nature of many well-performing methods poses a major obstacle for real-world deployment. Explainable AI (XAI) and especially feature attribution techniques promise to enable insights about how such models form their decision. But the plain application of such methods often fails to provide truly informative and problem-specific insights to domain experts. In this work, we focus on the specific task of detecting faults in rolling element bearings from vibration signals. We propose a novel and domain-specific feature attribution framework that allows us to evaluate how well the underlying logic of a model corresponds with expert reasoning. Utilizing the framework we are able to validate the trustworthiness and to successfully anticipate the generalization ability of different well-performing deep learning models. Our methodology demonstrates how signal processing tools can effectively be used to enhance Explainable AI techniques and acts as a template for similar problems.

cs.LG↗

arXiv2023

Explaining Deep Neural Networks for Bearing Fault Detection with Vibration Concepts

Concept-based explanation methods, such as Concept Activation Vectors, are potent means to quantify how abstract or high-level characteristics of input data influence the predictions of complex deep neural networks. However, applying them to industrial prediction problems is challenging as it is not immediately clear how to define and access appropriate concepts for individual use cases and specific data types. In this work, we investigate how to leverage established concept-based explanation techniques in the context of bearing fault detection with deep neural networks trained on vibration signals. Since bearings are prevalent in almost every rotating equipment, ensuring the reliability of intransparent fault detection models is crucial to prevent costly repairs and downtimes of industrial machinery. Our evaluations demonstrate that explaining opaque models in terms of vibration concepts enables human-comprehensible and intuitive insights about their inner workings, but the underlying assumptions need to be carefully validated first.

cs.LG↗

arXiv2023

Towards Scenario-based Safety Validation for Autonomous Trains with Deep Generative Models

Modern AI techniques open up ever-increasing possibilities for autonomous vehicles, but how to appropriately verify the reliability of such systems remains unclear. A common approach is to conduct safety validation based on a predefined Operational Design Domain (ODD) describing specific conditions under which a system under test is required to operate properly. However, collecting sufficient realistic test cases to ensure comprehensive ODD coverage is challenging. In this paper, we report our practical experiences regarding the utility of data simulation with deep generative models for scenario-based ODD validation. We consider the specific use case of a camera-based rail-scene segmentation system designed to support autonomous train operation. We demonstrate the capabilities of semantically editing railway scenes with deep generative models to make a limited amount of test data more representative. We also show how our approach helps to analyze the degree to which a system complies with typical ODD requirements. Specifically, we focus on evaluating proper operation under different lighting and weather conditions as well as while transitioning between them.

cs.LG↗

arXiv2023

The Thousand Faces of Explainable AI Along the Machine Learning Life Cycle: Industrial Reality and Current State of Research

In this paper, we investigate the practical relevance of explainable artificial intelligence (XAI) with a special focus on the producing industries and relate them to the current state of academic XAI research. Our findings are based on an extensive series of interviews regarding the role and applicability of XAI along the Machine Learning (ML) lifecycle in current industrial practice and its expected relevance in the future. The interviews were conducted among a great variety of roles and key stakeholders from different industry sectors. On top of that, we outline the state of XAI research by providing a concise review of the relevant literature. This enables us to provide an encompassing overview covering the opinions of the surveyed persons as well as the current state of academic research. By comparing our interview results with the current research approaches we reveal several discrepancies. While a multitude of different XAI approaches exists, most of them are centered around the model evaluation phase and data scientists. Their versatile capabilities for other stages are currently either not sufficiently explored or not popular among practitioners. In line with existing work, our findings also confirm that more efforts are needed to enable also non-expert users' interpretation and understanding of opaque AI models with existing methods and frameworks.

cs.LG↗

arXiv2017

Infrared properties of micromachined vanadium oxide thin films

This paper discusses questions of synthesizing and pressing vanadium oxides to create film-forming materials that can be used in producing optical coatings. Based on the film-forming materials thus created, technological processes have been developed for fabricating coatings from vanadium dioxide by two methods of producing thin films: vacuum evaporation and magnetron sputtering. Questions of using films made from vanadium oxide in optical instrumentation are considered.

cond-mat.mtrl-sci↗

arXiv2014

An efficient quantum algorithm for finding hidden parabolic subgroups in the general linear group

In the theory of algebraic groups, parabolic subgroups form a crucial building block in the structural studies. In the case of general linear groups over a finite field $F_q$, given a sequence of positive integers $n_1, ..., n_k$, where $n=n_1+...+n_k$, a parabolic subgroup of parameter $(n_1, ..., n_k)$ in $GL_n(F_q)$ is a conjugate of the subgroup consisting of block lower triangular matrices where the $i$th block is of size $n_i$. Our main result is a quantum algorithm of time polynomial in $\log q$ and $n$ for solving the hidden subgroup problem in $GL_n(F_q)$, when the hidden subgroup is promised to be a parabolic subgroup. Our algorithm works with no prior knowledge of the parameter of the hidden parabolic subgroup. Prior to this work, such an efficient quantum algorithm was only known for the case $n=2$ (A. Denney, C. Moore, and A. Russell (2010), Quantum Inf. Comput., Vol. 10, pp. 282-291), and for minimal parabolic subgroups (Borel subgroups), for the case when $q$ is not much smaller than $n$ (G. Ivanyos: Quantum Inf. Comput., Vol. 12, pp. 661-669).

quant-ph↗

arXiv2013

Polynomial time quantum algorithms for certain bivariate hidden polynomial problems

We present a new method for solving the hidden polynomial graph problem (HPGP) which is a special case of the hidden polynomial problem (HPP). The new approach yields an efficient quantum algorithm for the bivariate HPGP even when the input consists of several level set superpositions, a more difficult version of the problem than the one where the input is given by an oracle. For constant degree, the algorithm is polylogarithmic in the size of the base field. We also apply the results to give an efficient quantum algorithm for the oracle version of the HPP for an interesting family of bivariate hidden functions. This family includes diagonal quadratic forms and elliptic curves.

quant-ph↗

arXiv2012

Hidden Symmetry Subgroup Problems

We advocate a new approach of addressing hidden structure problems and finding efficient quantum algorithms. We introduce and investigate the Hidden Symmetry Subgroup Problem (HSSP), which is a generalization of the well-studied Hidden Subgroup Problem (HSP). Given a group acting on a set and an oracle whose level sets define a partition of the set, the task is to recover the subgroup of symmetries of this partition inside the group. The HSSP provides a unifying framework that, besides the HSP, encompasses a wide range of algebraic oracle problems, including quadratic hidden polynomial problems. While the HSSP can have provably exponential quantum query complexity, we obtain efficient quantum algorithms for various interesting cases. To achieve this, we present a general method for reducing the HSSP to the HSP, which works efficiently in several cases related to symmetries of polynomials. The HSSP therefore connects in a rather surprising way certain hidden polynomial problems with the HSP. Using this connection, we obtain the first efficient quantum algorithm for the hidden polynomial problem for multivariate quadratic polynomials over fields of constant characteristic. We also apply the new methods to polynomial function graph problems and present an efficient quantum procedure for constant degree multivariate polynomials over any field. This result improves in several ways the currently known algorithms.

quant-ph↗

arXiv2008

Efficient Quantum Algorithm for Identifying Hidden Polynomials

We consider a natural generalization of an abelian Hidden Subgroup Problem where the subgroups and their cosets correspond to graphs of linear functions over a finite field F with d elements. The hidden functions of the generalized problem are not restricted to be linear but can also be m-variate polynomial functions of total degree n>=2. The problem of identifying hidden m-variate polynomials of degree less or equal to n for fixed n and m is hard on a classical computer since Omega(sqrt{d}) black-box queries are required to guarantee a constant success probability. In contrast, we present a quantum algorithm that correctly identifies such hidden polynomials for all but a finite number of values of d with constant probability and that has a running time that is only polylogarithmic in d.

quant-ph↗

arXiv2008

The Optimal Single Copy Measurement for the Hidden Subgroup Problem

The optimization of measurements for the state distinction problem has recently been applied to the theory of quantum algorithms with considerable successes, including efficient new quantum algorithms for the non-abelian hidden subgroup problem. Previous work has identified the optimal single copy measurement for the hidden subgroup problem over abelian groups as well as for the non-abelian problem in the setting where the subgroups are restricted to be all conjugate to each other. Here we describe the optimal single copy measurement for the hidden subgroup problem when all of the subgroups of the group are given with equal a priori probability. The optimal measurement is seen to be a hybrid of the two previously discovered single copy optimal measurements for the hidden subgroup problem.

quant-ph↗

arXiv2007

How much is a quantum controller controlled by the controlled system?

We consider unitary transformations on a bipartite system A x B. To what extent entails the ability to transmit information from A to B the ability to transfer information in the converse direction? We prove a dimension-dependent lower bound on the classical channel capacity C(A<--B) in terms of the capacity C(A-->B) for the case that the bipartite unitary operation consists of controlled local unitaries on B conditioned on basis states on A. This can be interpreted as a statement on the strength of the inevitable backaction of a quantum system on its controller. If the local operations are given by the regular representation of a finite group G we have C(A-->B)=log |G| and C(A<--B)=log N where N is the sum over the degrees of all inequivalent representations. Hence the information deficit C(A-->B)-C(A<--B) between the forward and the backward capacity depends on the "non-abelianness" of the control group. For regular representations, the ratio between backward and forward capacities cannot be smaller than 1/2. The symmetric group S_n reaches this bound asymptotically. However, for the general case (without group structure) all bounds must depend on the dimensions since it is known that the ratio can tend to zero.

quant-ph↗

arXiv2007

Efficient Quantum Algorithm for Hidden Quadratic and Cubic Polynomial Function Graphs

We introduce the Hidden Polynomial Function Graph Problem as a natural generalization of an abelian Hidden Subgroup Problem (HSP) where the subgroups and their cosets correspond to graphs of linear functions over the finite field F_p. For the Hidden Polynomial Function Graph Problem the functions are not restricted to be linear but can also be multivariate polynomial functions of higher degree. For a fixed number of indeterminates and bounded total degree the Hidden Polynomial Function Graph Problem is hard on a classical computer as its black box query complexity is polynomial in p. In contrast, this problem can be reduced to a quantum state identification problem so that the resulting quantum query complexity does not depend on p. For univariate polynomials we construct a von Neumann measurement for distinguishing the states. We relate the success probability and the implementation of this measurement to certain classical problems involving polynomial equations. We present an efficient algorithm for hidden quadratic and cubic function graphs by establishing that the success probability of the measurement is lower bounded by a constant and that it can be implemented efficiently.

quant-ph↗