This paper aims to explain adversarial attacks in terms of how adversarial perturbations contribute to the attacking task. We estimate attributions of different image regions to the decrease of the attacking cost based on the Shapley value. We define and quantify interactions among adversarial perturbation pixels, and decompose the entire perturbation map into relatively independent perturbation components. The decomposition of the perturbation map shows that adversarially-trained DNNs have more perturbation components in the foreground than normally-trained DNNs. Moreover, compared to the normally-trained DNN, the adversarially-trained DNN have more components which mainly decrease the score of the true category. Above analyses provide new insights into the understanding of adversarial attacks.
This paper proposes a hypothesis for the aesthetic appreciation that aesthetic images make a neural network strengthen salient concepts and discard inessential concepts. In order to verify this hypothesis, we use multi-variate interactions to represent salient concepts and inessential concepts contained in images. Furthermore, we design a set of operations to revise images towards more beautiful ones. In experiments, we find that the revised images are more aesthetic than the original ones to some extent.
This is the Proceedings of ICML 2021 Workshop on Theoretic Foundation, Criticism, and Application Trend of Explainable AI. Deep neural networks (DNNs) have undoubtedly brought great success to a wide range of applications in computer vision, computational linguistics, and AI. However, foundational principles underlying the DNNs' success and their resilience to adversarial attacks are still largely missing. Interpreting and theorizing the internal mechanisms of DNNs becomes a compelling yet controversial topic. This workshop pays a special interest in theoretic foundations, limitations, and new application trends in the scope of XAI. These issues reflect new bottlenecks in the future development of XAI.
The reasonable definition of semantic interpretability presents the core challenge in explainable AI. This paper proposes a method to modify a traditional convolutional neural network (CNN) into an interpretable compositional CNN, in order to learn filters that encode meaningful visual patterns in intermediate convolutional layers. In a compositional CNN, each filter is supposed to consistently represent a specific compositional object part or image region with a clear meaning. The compositional CNN learns from image labels for classification without any annotations of parts or regions for supervision. Our method can be broadly applied to different types of CNNs. Experiments have demonstrated the effectiveness of our method.
In this paper, we rethink how a DNN encodes visual concepts of different complexities from a new perspective, i.e. the game-theoretic multi-order interactions between pixels in an image. Beyond the categorical taxonomy of objects and the cognitive taxonomy of textures and shapes, we provide a new taxonomy of visual concepts, which helps us interpret the encoding of shapes and textures, in terms of concept complexities. In this way, based on multi-order interactions, we find three distinctive signal-processing behaviors of DNNs encoding textures. Besides, we also discover the flexibility for a DNN to encode shapes is lower than the flexibility of encoding textures. Furthermore, we analyze how DNNs encode outlier samples, and explore the impacts of network architectures on interactions. Additionally, we clarify the crucial role of the multi-order interactions in real-world applications. The code will be released when the paper is accepted.
This paper aims to formulate the problem of estimating the optimal baseline values for the Shapley value in game theory. The Shapley value measures the attribution of each input variable of a complex model, which is computed as the marginal benefit from the presence of this variable w.r.t.its absence under different contexts. To this end, people usually set the input variable to its baseline value to represent the absence of this variable (i.e.the no-signal state of this variable). Previous studies usually determine the baseline values in an empirical manner, which hurts the trustworthiness of the Shapley value. In this paper, we revisit the feature representation of a deep model from the perspective of game theory, and define the multi-variate interaction patterns of input variables to define the no-signal state of an input variable. Based on the multi-variate interaction, we learn the optimal baseline value of each input variable. Experimental results have demonstrated the effectiveness of our method.
This paper aims to understand adversarial attacks and defense from a new perspecitve, i.e., the signal-processing behavior of DNNs. We novelly define the multi-order interaction in game theory, which satisfies six properties. With the multi-order interaction, we discover that adversarial attacks mainly affect high-order interactions to fool the DNN. Furthermore, we find that the robustness of adversarially trained DNNs comes from category-specific low-order interactions. Our findings provide more insights into and make a revision of previous understanding for the shape bias of adversarially learned features. Besides, the multi-order interaction can also explain the recoverability of adversarial examples.
In this study, we define interaction components of different orders between two input variables based on game theory. We further prove that interaction components of different orders satisfy several desirable properties.
This paper aims to explain deep neural networks (DNNs) from the perspective of multivariate interactions. In this paper, we define and quantify the significance of interactions among multiple input variables of the DNN. Input variables with strong interactions usually form a coalition and reflect prototype features, which are memorized and used by the DNN for inference. We define the significance of interactions based on the Shapley value, which is designed to assign the attribution value of each input variable to the inference. We have conducted experiments with various DNNs. Experimental results have demonstrated the effectiveness of the proposed method.