MACE: Model Agnostic Concept Extractor for Explaining Image Classification Networks

Kumar, Ashish; Sehgal, Karan; Garg, Prerna; Kamakshi, Vidhya; Chatapuramkrishnan, Narayanan

doi:10.1109/tai.2021.3111138

Cited by 9 publications

(7 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Word embedding technology is still developing and some pretrained language models are proposed to measure the similarity between words. Word2Vec was proposed in 2013 [ 8 ], GloVe in 2014 [ 9 ], OpenAIGPT in 2016 [ 10 ], ELMo (embeddings from language models) [ 11 , 12 ] and BERT (bidirectional encoder representation from transformers) were proposed in 2018 [ 13 , 14 ], and Transformer-XL [ 5 ] and XLNet [ 13 ] based on Transformers [ 15 ] architecture were proposed in 2019. The pretraining models commonly used in short-text affective orientation analysis include Word2Vec, GloVe, and BERT.…”

Section: Introductionmentioning

confidence: 99%

[Retracted] Emotion Analysis Based on Neural Network under the Big Data Environment

Zhou

Quan-ju

2022

Journal of Environmental and Public Health

View full text Add to dashboard Cite

Aiming at the problems of poor emotional tendency prediction effect and low utilization of syntactic information, this study proposes a big data sentiment analysis method based on neural network. First, the BERT model is used to vectorize the input data to reduce the semantic loss when the data is vectorized; then the word vector is input into the bidirectional LSTM encoder to obtain data features. Finally, the representation of the attention layer is used as the final feature vector for sentiment classification, reducing the influence of irrelevant data. The experimental results show that the method has high accuracy, recall, and F1 value and can effectively improve the accuracy of fine-grained sentiment classification of ambiguous texts.

show abstract

Section: Introductionmentioning

confidence: 99%

[Retracted] Emotion Analysis Based on Neural Network under the Big Data Environment

Zhou

Quan-ju

2022

Journal of Environmental and Public Health

View full text Add to dashboard Cite

show abstract

“…Wang et al [83] generate image patches [139] and use an attention mechanism to estimate the salient regions in a given image. However, a major limitation of these saliency map approaches is that they almost always highlight the region containing the entire object to be salient [38,63,64]. While these explanations can ascertain whether the model looks at the object to arrive at its prediction or relies on any non-object spurious correlations [32,34], finer explanations depicting the contributions of image primitives such as colors, textures, and parts cannot be obtained from the Class Activation Maps.…”

Section: Posthoc Methodsmentioning

confidence: 99%

“…Yeh et al [91] propose automatically extracting the complete set of concepts from the data, thereby preventing a possible loss of faithfulness due to leveraging concepts sampled from a different distribution [135]. Kumar et al [63] extend the capability of this framework [91] to unravel the complete blueprint of a class by formulating the concepts to be clustered in a class-specific fashion [52]. However, while extracting the explanations, these frameworks use multilayer nonlinear networks, which are also black boxes whose working could not be unraveled.…”

Section: Concept-based Explanationsmentioning

confidence: 99%

“…Union of regions comprising the concepts may be unfair to assess the goodness of the concepts as the union may cover up the entire image, nullifying the assessment. Recent works [63,64] propose a new metric called agreement accuracy which assesses how well the concept-based explainer approximates the working of the CNN to be explained. Leemann et al [173] propose using natural language models to assess the goodness of the concepts.…”

Section: Future Workmentioning

confidence: 99%

See 1 more Smart Citation

Explainable Image Classification: The Journey So Far and the Road Ahead

Kamakshi,

Krishnan

2023

Self Cite

View full text Add to dashboard Cite

Explainable Artificial Intelligence (XAI) has emerged as a crucial research area to address the interpretability challenges posed by complex machine learning models. In this survey paper, we provide a comprehensive analysis of existing approaches in the field of XAI, focusing on the tradeoff between model accuracy and interpretability. Motivated by the need to address this tradeoff, we conduct an extensive review of the literature, presenting a multi-view taxonomy that offers a new perspective on XAI methodologies. We analyze various sub-categories of XAI methods, considering their strengths, weaknesses, and practical challenges. Moreover, we explore causal relationships in model explanations and discuss approaches dedicated to explaining cross-domain classifiers. The latter is particularly important in scenarios where training and test data are sampled from different distributions. Drawing insights from our analysis, we propose future research directions, including exploring explainable allied learning paradigms, developing evaluation metrics for both traditionally trained and allied learning-based classifiers, and applying neural architectural search techniques to minimize the accuracy–interpretability tradeoff. This survey paper provides a comprehensive overview of the state-of-the-art in XAI, serving as a valuable resource for researchers and practitioners interested in understanding and advancing the field.

show abstract

“…ICE measures the importance of its class-wise concepts using TCAV. Other methods learn concept vectors and a mapping to feature space either for all classes simultaneously (ConceptSHAP [8]) or for each class separate (MACE [26], PACE [27]). Importantly, each method defines a custom measure for concept importance which is only applicable within the respective framework.…”

Section: Concept Similarities To Quantitatively Characterize the Simi...mentioning

confidence: 99%

Sparse Subspace Clustering for Concept Discovery (SSCCD)

Vielhaben¹,

Blücher²,

Strodthoff³

2022

Preprint

View full text Add to dashboard Cite

Concepts are key building blocks of higher level human understanding. Explainable AI (XAI) methods have shown tremendous progress in recent years, however, local attribution methods do not allow to identify coherent model behavior across samples and therefore miss this essential component. In this work, we study concept-based explanations and put forward a new definition of concepts as lowdimensional subspaces of hidden feature layers. We novelly apply sparse subspace clustering to discover these concept subspaces. Moving forward, we derive insights from concept subspaces in terms of localized input (concept) maps, show how to quantify concept relevances and lastly, evaluate similarities and transferability between concepts. We empirically demonstrate the soundness of the proposed Sparse Subspace Clustering for Concept Discovery (SSCCD) method for a variety of different image classification tasks. This approach allows for deeper insights into the actual model behavior that would remain hidden from conventional input-level heatmaps.

show abstract

MACE: Model Agnostic Concept Extractor for Explaining Image Classification Networks

Cited by 9 publications

References 20 publications

[Retracted] Emotion Analysis Based on Neural Network under the Big Data Environment

[Retracted] Emotion Analysis Based on Neural Network under the Big Data Environment

Explainable Image Classification: The Journey So Far and the Road Ahead

Sparse Subspace Clustering for Concept Discovery (SSCCD)

Contact Info

Product

Resources

About