B-cos Networks: Alignment is All We Need for Interpretability

Böhle, Moritz; Fritz, Mario; Schiele, Bernt

doi:10.1109/cvpr52688.2022.01008

Cited by 22 publications

(10 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The last row of the table shows the results for an adapted model of M-ResNet aimed at making it more explainable. Recently, Bohle et al [11] proposed a so-called B-Cos transform which, when interchanged with linear transforms of neural networks, increases the networks' explainability by promoting the alignment of weightinput during training. The alignment pressure on the weights ensures that the model computations align with task-relevant features and therefore become explainable.…”

Section: State-of-the-art Results Reproductionmentioning

confidence: 99%

“…In this work, we make use of the iNNvestigate library [1] which implements many explainability methods and provides a common interface to evaluate 19 white-box approaches including Layer-wise Relevance Propagation (LRP) [9] using 12 different rules. In addition, we also evaluate explanations generated by the recently proposed B-Cos network adjustment [11].…”

Section: Explainability Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Analyzing the real-world applicability of DGA classifiers

Drichel

Meyer

Schüppen

et al. 2020

Proceedings of the 15th International Conference on Availability, Reliability and Security

View full text Add to dashboard Cite

Separating benign domains from domains generated by DGAs with the help of a binary classifier is a well-studied problem for which promising performance results have been published. The corresponding multiclass task of determining the exact DGA that generated a domain enabling targeted remediation measures is less well studied. Selecting the most promising classifier for these tasks in practice raises a number of questions that have not been addressed in prior work so far. These include the questions on which traffic to train in which network and when, just as well as how to assess robustness against adversarial attacks. Moreover, it is unclear which features lead a classifier to a decision and whether the classifiers are real-time capable. In this paper, we address these issues and thus contribute to bringing DGA detection classifiers closer to practical use. In this context, we propose one novel classifier based on residual neural networks for each of the two tasks and extensively evaluate them as well as previously proposed classifiers in a unified setting. We not only evaluate their classification performance but also compare them with respect to explainability, robustness, and training and classification speed. Finally, we show that our newly proposed binary classifier generalizes well to other networks, is time-robust, and able to identify previously unknown DGAs. CCS CONCEPTS• Security and privacy → Intrusion detection systems; Malware and its mitigation; • Computing methodologies → Machine learning.

show abstract

Section: State-of-the-art Results Reproductionmentioning

confidence: 99%

Section: Explainability Methodsmentioning

confidence: 99%

Analyzing the real-world applicability of DGA classifiers

Drichel

Meyer

Schüppen

et al. 2020

Proceedings of the 15th International Conference on Availability, Reliability and Security

View full text Add to dashboard Cite

show abstract

“…As a bird's view analysis, there are two main distinctions between methods: Interpretable by-design architectures, and Post-Hoc explainability methods. The former searches to create algorithms that directly expose why a decision was made [2,3,5,8,22,35,53]. Our research study is based on the latter.…”

Section: Explainable Artificial Intelligencementioning

confidence: 99%

Diffusion Models for Counterfactual Explanations

Jeanneret

Simon

Jurie

2023

Computer Vision – ACCV 2022

View full text Add to dashboard Cite

This paper addresses the challenge of generating Counterfactual Explanations (CEs), involving the identification and modification of the fewest necessary features to alter a classifier's prediction for a given image. Our proposed method, Text-to-Image Models for Counterfactual Explanations (TIME), is a black-box counterfactual technique based on distillation. Unlike previous methods, this approach requires solely the image and its prediction, omitting the need for the classifier's structure, parameters, or gradients. Before generating the counterfactuals, TIME introduces two distinct biases into Stable Diffusion in the form of textual embeddings: the context bias, associated with the image's structure, and the class bias, linked to class-specific features learned by the target classifier. After learning these biases, we find the optimal latent code applying the classifier's predicted class token and regenerate the image using the target embedding as conditioning, producing the counterfactual explanation. Extensive empirical studies validate that TIME can generate explanations of comparable effectiveness even when operating within a black-box setting.

show abstract

“…Aligning all used features with human concepts is still difficult, albeit more feasible than without a SLDD-Model. Future work could use a more interpretable feature extractor like B-cos Networks (Böhle et al, 2022) to alleviate that problem. The method could be improved via the feature selection or improvements for small datasets (Reinders et al, 2022).…”

Section: Limitations and Future Workmentioning

confidence: 99%

“…Our proposed method generates the SLDD-Model by utilizing glm-saga (Wong et al, 2021) to compute a sparse linear classifier for selected features, which we then finetune to the sparse structure. We apply feature selection instead of a transformation to reduce the computational load and preserve the original semantics of the features, which can improve interpretability (Tao et al, 2015), especially if a more interpretable model like B-cos Networks (Böhle et al, 2022) is used. Additionally, we propose a novel loss function for more diverse features, which is especially relevant when one class depends on very few features, since using more redundant features limits the total information available for the decision.…”

Section: Introductionmentioning

confidence: 99%

Take 5: Interpretable Image Classification with a Handful of Features

Norrenbrock¹,

Rudolph²,

Rosenhahn³

2023

Preprint

View full text Add to dashboard Cite

Deep Neural Networks use thousands of mostly incomprehensible features to identify a single class, a decision no human can follow. We propose an interpretable sparse and low dimensional final decision layer in a deep neural network with measurable aspects of interpretability and demonstrate it on fine-grained image classification. We argue that a human can only understand the decision of a machine learning model, if the features are interpretable and only very few of them are used for a single decision. For that matter, the final layer has to be sparse and -to make interpreting the features feasible -low dimensional. We call a model with a Sparse Low-Dimensional Decision "SLDD-Model". We show that a SLDD-Model is easier to interpret locally and globally than a dense high-dimensional decision layer while being able to maintain competitive accuracy. Additionally, we propose a loss function that improves a model's feature diversity and accuracy. Our more interpretable SLDD-Model only uses 5 out of just 50 features per class, while maintaining 97 % to 100 % of the accuracy on four common benchmark datasets compared to the baseline model with 2048 features. Figure 1: Local explanation by our SLDD-Model: The two features used for the predicted class, emerged without additional supervision, are aligned with human interpretable attributes and localized (described in App. D) adequately. 36th Conference on Neural Information Processing Systems (NeurIPS 2022).

show abstract

B-cos Networks: Alignment is All We Need for Interpretability

Cited by 22 publications

References 33 publications

Analyzing the real-world applicability of DGA classifiers

Analyzing the real-world applicability of DGA classifiers

Diffusion Models for Counterfactual Explanations

Take 5: Interpretable Image Classification with a Handful of Features

Contact Info

Product

Resources

About