Explainable AI Methods - A Brief Overview

Holzinger, Andreas; Saranti, Anna; Molnar, Christoph; Biecek, Przemysław; Samek, Wojciech

doi:10.1007/978-3-031-04083-2_2

Cited by 171 publications

(123 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…6. There is an extensive body of work on non-formal XAI approaches to XAI [1,136,164,75,163,175,77,78,153]. 7.…”

Section: Formal Explainabilitymentioning

confidence: 99%

“…Similarly, one could consider exploiting non-formal model-agnostic explainers. There is a large body of work on non-formal modelagnostic XAI approaches [1,136,164,75,163,175,77,78,153]. Well-known examples include LIME [155], SHAP [116] and Anchor [156], for model-agnostic explanations, and sensitivity analysis [172] and LRP [17] in the case of sliency maps for neural networks.…”

Section: Datasetmentioning

confidence: 99%

See 1 more Smart Citation

On Tackling Explanation Redundancy in Decision Trees

Izza,

Ignatiev,

Marques-Silva

2022

Preprint

View full text Add to dashboard Cite

Decision trees (DTs) epitomize the ideal of interpretability of machine learning (ML) models. The interpretability of decision trees motivates explainability approaches by socalled intrinsic interpretability, and it is at the core of recent proposals for applying interpretable ML models in high-risk applications. The belief in DT interpretability is justified by the fact that explanations for DT predictions are generally expected to be succinct. Indeed, in the case of DTs, explanations correspond to DT paths. Since decision trees are ideally shallow, and so paths contain far fewer features than the total number of features, explanations in DTs are expected to be succinct, and hence interpretable. This paper offers both theoretical and experimental arguments demonstrating that, as long as interpretability of decision trees equates with succinctness of explanations, then decision trees ought not be deemed interpretable. The paper introduces logically rigorous path explanations and path explanation redundancy, and proves that there exist functions for which decision trees must exhibit paths with explanation redundancy that is arbitrarily larger than the actual path explanation. The paper also proves that only a very restricted class of functions can be represented with DTs that exhibit no explanation redundancy. In addition, the paper includes experimental results substantiating that path explanation redundancy is observed ubiquitously in decision trees, including those obtained using different tree learning algorithms, but also in a wide range of publicly available decision trees. The paper also proposes polynomial-time algorithms for eliminating path explanation redundancy, which in practice require negligible time to compute. Thus, these algorithms serve to indirectly attain irreducible, and so succinct, explanations for decision trees. Furthermore, the paper includes novel results related with duality and enumeration of explanations, based on using SAT solvers as witness-producing NP-oracles.

show abstract

“…6. There is an extensive body of work on non-formal XAI approaches to XAI [1,136,164,75,163,175,77,78,153]. 7.…”

Section: Formal Explainabilitymentioning

confidence: 99%

Section: Datasetmentioning

confidence: 99%

On Tackling Explanation Redundancy in Decision Trees

Izza,

Ignatiev,

Marques-Silva

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…All these possibilities direct medical research and practice in prominent directions [16]: more reliable and precise health analytics and predictive modeling [7], power data visualization techniques, tailored therapies, recommendations and interventions, personal user-friendly interfaces for communication [9] between different participants and stakeholders.…”

Section: Discussionmentioning

confidence: 99%

“…In typical medical systems various predictive models are generated to support personalized medical decisions. To make results of predictive models more understandable to doctors and other end-users recently XAI methods are using [9], and different ways of data visualization are applied.…”

Section: Emergent Artificial Intelligence Approaches For Supporting Q...mentioning

confidence: 99%

AI Approaches in Processing and Using Data in Personalized Medicine

Ivanović

Autexier

Kokkonidis

2022

Advances in Databases and Information Systems

View full text Add to dashboard Cite

In modern dynamic constantly developing society, more and more people suffer from chronic and serious diseases and doctors and patients need special and sophisticated medical and health support. Accordingly, prominent health stakeholders have recognized the importance of development of such services to make patients' life easier. Such support requires the collection of: huge amount of patients' complex data (clinical, environmental, nutritional, daily ac-tivities…), variety of data from smart wearable devices, data from clothing equipped with sensors etc. Holistic patient's data must be properly aggregated, processed, analyzed, and presented to the doctors/caregivers to recommend adequate treatment and actions to improve patient's health related parameters and general wellbeing. Advanced artificial intelligence techniques offer the opportunity to analyze such big data, consume them, and derive new knowledge to support (personalized) medical decisions. New approaches like those based on advanced machine/deep learning, federated learning, transfer learning, explainable artificial intelligence open new paths for more quality use of health and medical data in future. In this paper, we will present some crucial aspects and characteristic examples in the area of application of a range of artificial intelligence approaches in (personalized) medical decisions.

show abstract

“…As a result, we have seen a suit of interpretable ML or X-AI methods particularly to untangle deep learning models [ 34 ]. One can choose from various ML interpretability techniques (shown in Figure 1 ) for any use case [ 34 , 35 , 36 , 37 ].…”

Section: Current State-of-the-art Explainable Ai Methods and Approachesmentioning

confidence: 99%

Explainable AI: A Neurally-Inspired Decision Stack Framework

Khan

Nayebpour

et al. 2022

Biomimetics

View full text Add to dashboard Cite

European law now requires AI to be explainable in the context of adverse decisions affecting the European Union (EU) citizens. At the same time, we expect increasing instances of AI failure as it operates on imperfect data. This paper puts forward a neurally inspired theoretical framework called “decision stacks” that can provide a way forward in research to develop Explainable Artificial Intelligence (X-AI). By leveraging findings from the finest memory systems in biological brains, the decision stack framework operationalizes the definition of explainability. It then proposes a test that can potentially reveal how a given AI decision was made.

show abstract

Explainable AI Methods - A Brief Overview

Cited by 171 publications

References 54 publications

On Tackling Explanation Redundancy in Decision Trees

On Tackling Explanation Redundancy in Decision Trees

AI Approaches in Processing and Using Data in Personalized Medicine

Explainable AI: A Neurally-Inspired Decision Stack Framework

Contact Info

Product

Resources

About