Many high-performing machine learning models are not interpretable. As they are increasingly used in decision scenarios that can critically affect individuals, it is necessary to develop tools to better understand the...
详细信息
ISBN:
(纸本)9783031440632;9783031440649
Many high-performing machine learning models are not interpretable. As they are increasingly used in decision scenarios that can critically affect individuals, it is necessary to develop tools to better understand their outputs. Popular explanation methods include contrastive explanations. However, they suffer several shortcomings, among others an insufficient incorporation of background knowledge, and a lack of interactivity. While (dialogue-like) interactivity is important to better communicate an explanation, background knowledge has the potential to significantly improve their quality, e.g., by adapting the explanation to the needs of the end-user. To close this gap, we present reasonx, an explanation tool based on constraint logic programming (CLP). reasonx provides interactive contrastive explanations that can be augmented by background knowledge, and allows to operate under a setting of under-specified information, leading to increased flexibility in the provided explanations. reasonx computes factual and contrastive decision rules, as well as closest contrastive examples. It provides explanations for decision trees, which can be the ML models under analysis, or global/local surrogate models of any ML model. While the core part of reasonx is built on CLP, we also provide a program layer that allows to compute the explanations via Python, making the tool accessible to a wider audience. We illustrate the capability of reasonx on a synthetic data set, and on a well-developed example in the credit domain. In both cases, we can show how reasonx can be flexibly used and tailored to the needs of the user.
This work presents the performance comparison of two conceptually different approaches for a mixed model non-permutation flowshop production line. The demand is a semi-dynamic demand with a fixed job sequence for the ...
详细信息
ISBN:
(纸本)9783540788256
This work presents the performance comparison of two conceptually different approaches for a mixed model non-permutation flowshop production line. The demand is a semi-dynamic demand with a fixed job sequence for the first station. Resequencing is permitted where stations have access to intermediate or centralized resequencing buffers. The access to the buffers is restricted by the number of available buffer places and the physical size of the products. An exact approach, using constraint logic programming (CLP), and a heuristic approach, a Genetic Algorithm (GA), were applied.
This paper introduces a unified constraint-based test case generator for white-box method-level unit testing. The derivation of a suite of test cases can be defined as a constraint satisfaction problem. Each test case...
详细信息
ISBN:
(纸本)9781509034383
This paper introduces a unified constraint-based test case generator for white-box method-level unit testing. The derivation of a suite of test cases can be defined as a constraint satisfaction problem. Each test case consists of a test input and an expected output. The program is automatically transformed into a constraint model called constraintlogic graph. The constraintlogic graph is a succinct graphical representation of the system of constraints that defines the relationships between the test inputs and actual outputs. The suite of test inputs can be solved from the conjunction of constraints on each complete path of the constraintlogic graph by a constraint logic programming language. The specification of a method is defined by the Object constraint Language. This non-executable specification is automatically transformed into an executable specification defined by a constraint logic programming language. This executable specification serves as the test oracle to automatically generate the corresponding expected output for a given test input.
Qualitative spatial reasoning (QSR) is useful for deriving logical inferences when quantitative spatial information is not available. QSR theories have applications in areas such as geographic information systems, spa...
详细信息
Qualitative spatial reasoning (QSR) is useful for deriving logical inferences when quantitative spatial information is not available. QSR theories have applications in areas such as geographic information systems, spatial databases, robotics, and cognitive sciences. The existing QSR theories have been applied primarily to 2D. The ability to perform QSR over a collection of 3D objects is desirable in many problem domains. Here we present the evolution (VRCC-3D+) of RCC-based QSR from 2D to both 3D (including occlusion support) and 4D (a temporal component). It is time consuming to construct large composition tables manually. We give a divideand- conquer algorithm to construct a comprehensive composition table from smaller constituent tables (which can be easily handcrafted). In addition to the logical consistency entailment checking that is required for such a system, clearly there is a need for a spatio-temporal component to account for spatial movements and path consistency (i.e. to consider only smooth transitions in spatial movements over time). Visually, these smooth movement phenomena are represented as a conceptual neighborhood graph. We believe that the methods presented herein to detect consistency, refine uncertainty, and enhance reasoning about 3D objects will provide useful guidelines for other studies in automated spatial reasoning.
This paper presents a universal model transformation method for the problem which is represented in the form of facts. This method is the key element of the hybrid approach. In this approach, two environments of const...
详细信息
ISBN:
(纸本)9783319672298;9783319672281
This paper presents a universal model transformation method for the problem which is represented in the form of facts. This method is the key element of the hybrid approach. In this approach, two environments of constraint logic programming (CLP) and mathematical programming (MP) were integrated and hybridized. Hybrid approach is very effective method for modeling and solving different kinds of decision and optimization problems in manufacturing, transportation, logistics etc. These problems are characterized by large numbers of integer variables and constraints, which build a large space of possible solutions. The model transformation method makes it possible to significantly reduce this space even before looking for a solution. The paper also presents an author's illustrative model for variant of CVRP (Capacity Vehicle Routing Problem) and many numerical experiments to test the model transformation method and evaluate its effectiveness when using a hybrid approach.
We refine the mathematical specification of a WAM extension to type-constraint logic programming given in [BeB96]. We provide a full specification and correctness proof of the PROTOS Abstract Machine (PAM), an extensi...
详细信息
We refine the mathematical specification of a WAM extension to type-constraint logic programming given in [BeB96]. We provide a full specification and correctness proof of the PROTOS Abstract Machine (PAM), an extension of the WAM by polymorphic order-sorted unification as required by the logicprogramming language PROTOS-L, by refining the abstract type constraints used in [BeB96] to the polymorphic order-sorted types of PROTOS-L. This allows us to develop a detailed and mathematically precise account of the PAM's compiled type constraint representation and solving facilities, and to extend the correctness theorem to compilation on the fully specified PAM.
We study contractibility and its approximation for two very general classes of soft global constraints. We introduce a general formulation of decomposition-based soft constraints and provide a sufficient condition for...
详细信息
ISBN:
(纸本)9783939897170
We study contractibility and its approximation for two very general classes of soft global constraints. We introduce a general formulation of decomposition-based soft constraints and provide a sufficient condition for contractibility and an approach to approximation. For edit-based soft constraints, we establish that the tightest contractible approximation cannot be expressed in edit-based terms, in general.
The technique of Abstract Interpretation [13] has allowed the development of sophisticated program analyses which are provably correct and practical. The semantic approximations produced by such analyses have been tra...
详细信息
ISBN:
(数字)9783540366072
ISBN:
(纸本)3540009868
The technique of Abstract Interpretation [13] has allowed the development of sophisticated program analyses which are provably correct and practical. The semantic approximations produced by such analyses have been traditionally applied to optimization during program compilation. However, recently, novel and promising applications of semantic approximations have been proposed in the more general context of program verification and debugging [3],[10],[7].
Augmented shared workspaces are an instance of augmented reality which move the collaborative work from the desktop to the real workplace, enabling higher interaction level with coworkers and allowing implementation o...
详细信息
ISBN:
(纸本)0769513301
Augmented shared workspaces are an instance of augmented reality which move the collaborative work from the desktop to the real workplace, enabling higher interaction level with coworkers and allowing implementation of advanced techniques of human-machine interfaces. In this paper, we advocate the employment of a constraint system as the document model (or program) for the implementation of augmented reality applications. We show that constraint technology is synergistic with augmented reality and collaborative work in many aspects. An architecture model based on agents and constraints is then proposed. Our point of view is supported by the observation of this technology being employed in Computer Graphics and Human-Computer Interaction research and the implementation of typical cases.
Functional constraints and bi-functional constraints are an important constraint class in constraintprogramming (CP) systems, in particular for constraint logic programming (CLP) systems. CP systems with finite domai...
详细信息
Functional constraints and bi-functional constraints are an important constraint class in constraintprogramming (CP) systems, in particular for constraint logic programming (CLP) systems. CP systems with finite domain constraints usually employ constraint Satisfaction Problem(s)-based solvers which use local consistency, for example, arc consistency. We introduce a new approach which is based instead on variable substitution. We obtain efficient algorithms for reducing systems involving functional and bi-functional constraints together with other nonfunctional constraints. It also solves globally any CSP where there exists a variable such that any other variable is reachable from it through a sequence of functional constraints. Our experiments on random problems show that variable elimination can significantly improve the efficiency of solving problems with functional constraints.
暂无评论