Kandinsky Patterns

Mueller, Heimo; Holzinger, Andreas

doi:10.1016/j.artint.2021.103546

Cited by 15 publications

(5 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Kandinsky Pattern data sets are artificially created to test the reasoning and generalization capabilities of state-of-the-art neural networks [ 54 ]. We chose to use a Kandinsky Pattern data set as a benchmark, since the structure of some patterns is inherently relational; therefore they were first introduced in the medical domain, specifically in the classification and understanding of histopathological images [ 55 ].…”

Section: A Kandinsky Pattern Benchmark For Conceptual Validationmentioning

confidence: 99%

“…The objects follow certain constraints; specifically, they do not overlap, are not cropped at the border of the image, and are easily recognizable and clearly distinguishable by a human observer. Given a domain of object shapes, colors, sizes, and positions, a Kandinsky Pattern (KP) is then the subset of all possible KFs in this domain that adhere to a certain ground truth [ 54 ]. An example of such a ground truth could be a natural language statement, such as “The figure is vertically symmetrical”.…”

Section: A Kandinsky Pattern Benchmark For Conceptual Validationmentioning

confidence: 99%

“…First, we introduce our concept learning data set in Sect. 2 , in particular the use case of symmetry in Kandinsky Patterns [ 54 ] used to benchmark our ILP-based proof-of-concept implementation. We state which outcomes the benchmark should produce accordingly.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Generating Explanations for Conceptual Validation of Graph Neural Networks: An Investigation of Symbolic Predicates Learned on Relevance-Ranked Sub-Graphs

et al. 2022

Self Cite

View full text Add to dashboard Cite

Graph Neural Networks (GNN) show good performance in relational data classification. However, their contribution to concept learning and the validation of their output from an application domain’s and user’s perspective have not been thoroughly studied. We argue that combining symbolic learning methods, such as Inductive Logic Programming (ILP), with statistical machine learning methods, especially GNNs, is an essential forward-looking step to perform powerful and validatable relational concept learning. In this contribution, we introduce a benchmark for the conceptual validation of GNN classification outputs. It consists of the symbolic representations of symmetric and non-symmetric figures that are taken from a well-known Kandinsky Pattern data set. We further provide a novel validation framework that can be used to generate comprehensible explanations with ILP on top of the relevance output of GNN explainers and human-expected relevance for concepts learned by GNNs. Our experiments conducted on our benchmark data set demonstrate that it is possible to extract symbolic concepts from the most relevant explanations that are representative of what a GNN has learned. Our findings open up a variety of avenues for future research on validatable explanations for GNNs.

show abstract

Section: A Kandinsky Pattern Benchmark For Conceptual Validationmentioning

confidence: 99%

Section: A Kandinsky Pattern Benchmark For Conceptual Validationmentioning

confidence: 99%

See 1 more Smart Citation

Generating Explanations for Conceptual Validation of Graph Neural Networks: An Investigation of Symbolic Predicates Learned on Relevance-Ranked Sub-Graphs

et al. 2022

Self Cite

View full text Add to dashboard Cite

show abstract

“…The use of Logic Tensor Networks [22] could allow to improve the model by using real logic beyond unique is-a relationships. Also, it would be interesting to test this framework on a new benchmark dataset specifically containing images that are labeled with position of objects, object groupings and relational concept such as those involved in solving the challenges of Kandinsky Patterns [103,104] Finally, let's take look at the Renaissance line in the Figure 5. We see that the attributes are not very discriminative in favor of this class: the highest positive value is a weight of 0.14 for the attribute Serliana.…”

Section: Discussionmentioning

confidence: 99%

Greybox XAI: A Neural-Symbolic learning framework to produce interpretable predictions for image classification

Bennetot¹,

Franchi²,

Ser³

et al. 2022

Knowledge-Based Systems

View full text Add to dashboard Cite

“…For example, Shrikumar et al (2017) perform an experiment evaluating different explanation methods for image classification based on the decrease in the classification accuracy on the MNIST dataset (LeCun et al, 2010) after masking features identified as important by an explanation method. Another example is the dataset of Kandinsky Patterns and accompanying challenges introduced by M✓ller and Holzinger (2021): in brief, challenges comprise classifying simple visual patterns in controllable synthetic image datasets while producing explanations in a specific format, for example, natural language. Similarly, by extending CLEVR dataset (Johnson et al, 2017) for visual question answering, Arras et al (2020) release the CLEVR‐XAI benchmark for neural network explanation methods.…”

Section: Evaluation Of Interpretability and Explainabilitymentioning

confidence: 99%

Interpretable and explainable machine learning: A methods‐centric overview with concrete examples

Marcinkevičs

Vogt

2023

WIREs Data Min & Knowl

View full text Add to dashboard Cite

Interpretability and explainability are crucial for machine learning (ML) and statistical applications in medicine, economics, law, and natural sciences and form an essential principle for ML model design and development. Although interpretability and explainability have escaped a precise and universal definition, many models and techniques motivated by these properties have been developed over the last 30 years, with the focus currently shifting toward deep learning. We will consider concrete examples of state-of-the-art, including specially tailored rule-based, sparse, and additive classification models, interpretable representation learning, and methods for explaining black-box models post hoc. The discussion will emphasize the need for and relevance of interpretability and explainability, the divide between them, and the inductive biases behind the presented "zoo" of interpretable models and explanation methods.

show abstract

Kandinsky Patterns

Cited by 15 publications

References 22 publications

Generating Explanations for Conceptual Validation of Graph Neural Networks: An Investigation of Symbolic Predicates Learned on Relevance-Ranked Sub-Graphs

Generating Explanations for Conceptual Validation of Graph Neural Networks: An Investigation of Symbolic Predicates Learned on Relevance-Ranked Sub-Graphs

Greybox XAI: A Neural-Symbolic learning framework to produce interpretable predictions for image classification

Interpretable and explainable machine learning: A methods‐centric overview with concrete examples

Contact Info

Product

Resources

About