1,721,088 research outputs found
ExACT Explainable Clustering: Unravelling the Intricacies of Cluster Formation
No full textCluster assignments, in particular the deep clustering ones, are often hard to explain, partially because they depend on all the features of the data in a complicated way, so it is difficult to determine why a particular row of data is classified in a particular bucket. This opaqueness makes their predictions not trustable, as for many predictors based on black boxes. This paper aims to tackle the aforementioned issues by introducing the design and implementation of ExACT, a new explainable clustering algorithm based on the induction of decision trees and performing hypercubic approximations of the input feature space in order to provide output human-interpretable clusters. Furthermore, ExACT is versatile enough to perform explainable classification and regression as well, as demonstrated in this work, proving to be a competitive alternative to existing analogous algorithms
Untying black boxes with clustering-based symbolic knowledge extraction
Full text linkMachine learning black boxes, exemplified by deep neural networks, often exhibit challenges in interpretability due to their reliance on complicated relationships involving numerous internal parameters and input features. This lack of transparency from a human perspective renders their predictions untrustworthy, particularly in critical applications. In this paper, we address this issue by introducing the design and implementation of CReEPy, an algorithm for symbolic knowledge extraction based on explainable clustering. Specifically, CReEPy leverages the underlying clustering performed by the ExACT or CREAM algorithms to generate human-interpretable Prolog rules that mimic the behaviour of opaque models. Additionally, we introduce CRASH, an algorithm for the automated tuning of hyper-parameters required by CReEPy. We present experiments evaluating both the human readability and predictive performance of the proposed knowledge-extraction algorithm, employing existing state-of-the-art techniques as benchmarks for comparison in real-world applications
Achieving Complete Coverage with Hypercube-Based Symbolic Knowledge-Extraction Techniques
Full text linkSymbolic knowledge-extraction (SKE) techniques are currently employed for various purposes, particularly addressing the challenge of explaining opaque models by generating human-understandable explanations. The existing literature encompasses a diverse range of techniques, each relying on specific theoretical assumptions and possessing its own advantages and disadvantages. Amongst the available choices, hypercube-based SKE techniques are notable for their adaptability and versatility. However, they may suffer from limited completeness when utilised for making predictions. This research aims to augment the predictive capabilities of hypercube-based SKE techniques, striving for a completeness rate of 100%. Furthermore, the study includes experiments that assess the effectiveness of the proposed enhancements
Unlocking Insights and Trust: The Value of Explainable Clustering Algorithms for Cognitive Agents
Full text linkIn the realm of cognitive agents, including both human users and AI systems, explainable clustering algorithms have gained prominence. These algorithms offer enhanced transparency, making clustering results comprehensible to users and aiding AI systems in decision-making. They also facilitate knowledge discovery by revealing cluster characteristics, reducing cognitive load for users, and playing a vital role in ethical and bias mitigation. This paper introduces an innovative extension of the existing PSyKE framework, designed to support explainable clustering techniques and, thus, to augment cognitive agent capabilities. State-of-the-art review, experiment findings, and a synthesis of key insights are also provided
Hierarchical Knowledge Extraction from Opaque Machine Learning Predictors
No full textAdopting opaque machine learning predictors, which achieve very high predictive performance, often necessitates incorporating symbolic knowledge-extraction techniques. These techniques aim to explain the opaque predictions, thus making them applicable in high-stakes scenarios. The development of symbolic knowledge-extraction procedures is evolving alongside the dynamic machine learning landscape. However, there are recurring drawbacks that tend to be overlooked or addressed in a suboptimum way. Common examples include the non-exhaustiveness of the global explanations generated for a black-box predictor or the unwanted discretisation introduced in the prediction of continuous variables. To tackle these challenges, in this work, we introduce the HEx algorithm, its formalisation and its properties. This algorithm aims to obtain a symbolic, hierarchical representation of the knowledge acquired by opaque machine learning classifiers and regressors, always ensuring knowledge exhaustiveness and avoiding any output discretisation. Experiments demonstrating the superior capabilities of HEx compared to state-of-the-art competitors in terms of predictive performance, completeness, and human readability are presented
Unveiling Opaque Predictors via Explainable Clustering: The CReEPy Algorithm
Full text linkMachine learning black boxes, as deep neural networks, are often hard to explain because their predictions depend on complicated relationships involving a huge amount of internal parameters and input features. This opaqueness from the human perspective makes their predictions not trustable, especially in critical applications. In this paper we tackle this issue by introducing the design and implementation of CReEPy, an algorithm performing symbolic knowledge extraction based on explainable clustering. In particular, CReEPy relies on the underlying clustering performed by the ExACT or CREAM procedures to provide human-interpretable Prolog rules mimicking the behaviour of the opaque model. Experiments to assess both the human readability and the predictive performance of the proposed algorithm are discussed here, using existing state-of-the-art techniques as benchmarks for the comparison
Unmasking the Shadows: Leveraging Symbolic Knowledge Extraction to Discover Biases and Unfairness in Opaque Predictive Models
Full text linkThis work explores the efficacy of symbolic knowledge-extraction (SKE) techniques in identifying biases and unfairness within opaque predictive models. Logic rules extracted from black-box predictors make it possible to verify if decisions are influenced by protected or sensitive features. In particular, the identification of biased or unfair decisions can be achieved through the evaluation of if-then rules, detecting the inclusion of protected and/or sensitive information in the rules’ precondition. The effectiveness of SKE in this regard is demonstrated here by conducting various simulations on a well-known data set for loan grant prediction. Our findings highlight the potential of SKE as a valuable tool to reveal biases and discrimination in opaque predictions, ultimately contributing to the pursuit of fair and transparent decision-making systems
Bottom-Up and Top-Down Workflows for Hypercube- And Clustering-Based Knowledge Extractors
No full textMachine learning opaque models, currently exploited to carry out a wide variety of supervised and unsupervised learning tasks, are able to achieve impressive predictive performances. However, they act as black boxes (BBs) from the human standpoint, so they cannot be entirely trusted in critical applications unless there exists a method to extract symbolic and human-readable knowledge out of them. In this paper we analyse a recurrent design adopted by symbolic knowledge extractors for BB predictors—that is, the creation of rules associated with hypercubic input space regions. We argue that this kind of partitioning may lead to suboptimum solutions when the data set at hand is sparse, high-dimensional, or does not satisfy symmetric constraints. We then propose two different knowledge-extraction workflows involving clustering approaches, highlighting the possibility to outperform existing knowledge-extraction techniques in terms of predictive performance on data sets of any kind
Study of the Steam Turbine Trip in a 20 MW Geothermal Plant
No full textIn vapour-dominated geothermal sources a large amount of non-condensable gas is mixed with the steam and requires extraction from the condenser, generally by means of a compressor. A plant has been studied in which this compressor is directly coupled, via gearboxes, to the steam turbine shaft. When a 'trip' occurs (and this may often happen when considerable load variations cause a small plant to get out of synchronism), very high torque readings have been measured at the turbine-compressor coupling. This paper presents a computer simulation which has been developed for the system to explore ways of reducing these high torque
- …
