Search

Advancing defensive mechanisms against adversarial attacks in generative models is a critical research topic in machine learning. Our study focuses on a specific type of generative models - Variational Auto-Encoders (VAEs). Contrary to common beliefs and existing literature which suggest that noise injection towards training data can make models more robust, our preliminary experiments revealed that naive usage of noise augmentation technique did not substantially improve VAE robustness. In fact, it even degraded the quality of learned representations, making VAEs more susceptible to adversarial perturbations. This paper introduces a novel framework that enhances robustness by regularizing the latent space divergence between original and noise-augmented data. Through incorporating a paired probabilistic prior into the standard variational lower bound, our method significantly boosts defense against adversarial attacks. Our empirical evaluations demonstrate that this approach, termed Robust Augmented Variational Auto-ENcoder (RAVEN), yields superior performance in resisting adversarial inputs on widely-recognized benchmark datasets.

This paper proposes a new algorithm for learning accurate tree-based models while ensuring the existence of recourse actions. Algorithmic Recourse (AR) aims to provide a recourse action for altering the undesired prediction result given by a model. Typical AR methods provide a reasonable action by solving an optimization task of minimizing the required effort among executable actions. In practice, however, such actions do not always exist for models optimized only for predictive performance. To alleviate this issue, we formulate the task of learning an accurate classification tree under the constraint of ensuring the existence of reasonable actions for as many instances as possible. Then, we propose an efficient top-down greedy algorithm by leveraging the adversarial training techniques. We also show that our proposed algorithm can be applied to the random forest, which is known as a popular framework for learning tree ensembles. Experimental results demonstrated that our method successfully provided reasonable actions to more instances than the baselines without significantly degrading accuracy and computational efficiency., Comment: 27 pages, 10 figures, to appear in the 41st International Conference on Machine Learning (ICML 2024)

Achieving aesthetically pleasing photography necessitates attention to multiple factors, including composition and capture conditions, which pose challenges to novices. Prior research has explored the enhancement of photo aesthetics post-capture through 2D manipulation techniques; however, these approaches offer limited search space for aesthetics. We introduce a pioneering method that employs 3D operations to simulate the conditions at the moment of capture retrospectively. Our approach extrapolates the input image and then reconstructs the 3D scene from the extrapolated image, followed by an optimization to identify camera parameters and image aspect ratios that yield the best 3D view with enhanced aesthetics. Comparative qualitative and quantitative assessments reveal that our method surpasses traditional 2D editing techniques with superior aesthetics., Comment: 10 pages

In this work, we introduce two types of makeup prior models to extend existing 3D face prior models: PCA-based and StyleGAN2-based priors. The PCA-based prior model is a linear model that is easy to construct and is computationally efficient. However, it retains only low-frequency information. Conversely, the StyleGAN2-based model can represent high-frequency information with relatively higher computational cost than the PCA-based model. Although there is a trade-off between the two models, both are applicable to 3D facial makeup estimation and related applications. By leveraging makeup prior models and designing a makeup consistency module, we effectively address the challenges that previous methods faced in robustly estimating makeup, particularly in the context of handling self-occluded faces. In experiments, we demonstrate that our approach reduces computational costs by several orders of magnitude, achieving speeds up to 180 times faster. In addition, by improving the accuracy of the estimated makeup, we confirm that our methods are highly advantageous for various 3D facial makeup applications such as 3D makeup face reconstruction, user-friendly makeup editing, makeup transfer, and interpolation., Comment: CVPR2024. Project: https://yangxingchao.github.io/makeup-priors-page

We study the $U(1)_A$ anomaly at high temperatures of $N_f=2+1$ lattice QCD with chiral fermions. Gauge ensembles are generated with M\"obius domain-wall (MDW) fermions, and the measurements are reweighted to those with overlap fermions. We report on the results for the Dirac spectra, the $U(1)_A$ susceptibility, and the topological susceptibility in the temperature range of $T=136$, $153$, $175$, and $204$ MeV, where the up and down quark masses are set to be near the physical point as well as at lighter or heavier masses., Comment: 10 pages, 4 figures, 1 table; Proceedings of the 40th International Symposium on Lattice Field Theory (Lattice 2023), July 31st - August 4th, 2023, Fermilab, Batavia, Illinois, USA

Based on simulations of 2+1 flavor lattice QCD with M\"obius domain wall fermions at high temperatures, we compute a series of spatial correlation functions to study the screening masses in mesonic states. We compare these masses with the symmetry relations for various quark masses and lattice sizes at temperatures above the critical point. Using these spatial correlation functions we examine the $SU(2)_L \times SU(2)_R$ symmetry as well as the anomalously broken axial $U(1)_A$ symmetry. Additionally we explore a possible and emergent chiral-spin symmetry $SU(2)_{CS}$., Comment: 10 pages, 5 figures, Contribution to the 40th International Symposium on Lattice Field Theory (Lattice 2023)

We investigate relations between the chiral susceptibility and axial $U(1)$ anomaly in lattice QCD at high temperatures. Employing the exactly chiral symmetric Dirac operator, we separate the purely axial $U(1)$ breaking effect in the connected and disconnected chiral susceptibilites in a theoretically clean manner. Preliminary results for two-flavor lattice QCD near the critical temperature are presented., Comment: 9 pages, 3 figures, LATTICE2023, minor corrections

We present an update on the study of the QCD phase transition with 3 flavors of M\"obius domain wall fermions at zero baryon density. We performed simulations on lattices of size $36^3\times12\times16$ and $24^3\times12\times32$ with a variety of quark masses at a fixed lattice spacing $a=0.1361(20)$ fm, which correspond to a temperature 121(2) MeV. By analyzing the chiral condensate, chiral susceptibilitities and Binder cumulant on $36^3\times12\times16$ lattices together with the result obtained from our previous study on $24^3\times12\times16$ lattices, we identified a crossover occurring at quark mass around $m_q^{\mathrm{\overline {MS}}}(2\, \mathrm{GeV}) \sim 3-4$ MeV for this temperature. Besides, we show the effects of residual chiral symmetry breaking on chiral condensate and chiral susceptibilities between $L_s=16$ and 32., Comment: 10 pages, 6 figures, Contribution to the 40th International Symposium on Lattice Field Theory (Lattice 2023), July 31 to August 4, 2023, Fermilab, USA

Pose and body shape editing in a human image has received increasing attention. However, current methods often struggle with dataset biases and deteriorate realism and the person's identity when users make large edits. We propose a one-shot approach that enables large edits with identity preservation. To enable large edits, we fit a 3D body model, project the input image onto the 3D model, and change the body's pose and shape. Because this initial textured body model has artifacts due to occlusion and the inaccurate body shape, the rendered image undergoes a diffusion-based refinement, in which strong noise destroys body structure and identity whereas insufficient noise does not help. We thus propose an iterative refinement with weak noise, applied first for the whole body and then for the face. We further enhance the realism by fine-tuning text embeddings via self-supervised learning. Our quantitative and qualitative evaluations demonstrate that our method outperforms other existing methods across various datasets., Comment: Accepted to WACV 2024, project page: https://www.cgg.cs.tsukuba.ac.jp/~okuyama/pub/diffbody/

The TMD soft function can be obtained by formulating the Wilson line in terms of auxiliary 1-dimensional fermion fields on the lattice. In this formulation, the directional vector of the auxiliary field in Euclidean space has the form $\tilde n = (in^0, \vec 0_\perp, n^3)$, where the time component is purely imaginary. The components of these complex directional vectors in the Euclidean space can be mapped directly to the rapidities of the Minkowski space soft function. We present the results of the one-loop calculation of the Euclidean space analog to the soft function using these complex directional vectors. As a result, we show that the calculation is valid only when the directional vectors obey the relation: $|r| = |n^3/n^0| > 1$, and that this result corresponds to a computation in Minkowski space with space-like directed Wilson lines. Finally, we show that a lattice calculable object can be constructed that has the desired properties of the soft function., Comment: The 40th International Symposium on Lattice Field Theory (Lattice 2023)

The light-cone distribution amplitude (LCDA) of the pion contains information about the parton momentum carried by the quarks and is an important theoretical input for various predictions of exclusive processes at high energy, including the pion electromagnetic form factor. Progress towards constraining the fourth Mellin moment of the LCDA using the heavy-quark operator product expansion (HOPE) method is presented., Comment: 10 pages, 4 figures, proceedings to talk presented at the 40th International Symposium on Lattice Field Theory, July 31st - August 4th 2023, Fermi National Accelerator Laboratory, Batavia, Illinois, USA. arXiv admin note: text overlap with arXiv:2211.17009

While vehicles have primarily been controlled through mechanical means in years past, an increasing number of embedded control systems are being installed and used, keeping pace with advances in electronic control technology and performance. Automotive systems consist of multiple components developed by a range of vendors. To accelerate developments in embedded control systems, industrial standards such as AUTOSAR are being defined for automotive systems, including the design of operating system and middleware technologies. Crucial to ensuring the safety of automotive systems, the operating system is foundational software on which many automotive applications are executed. In this paper, we propose an integrated model-based method for verifying automotive operating systems; our method is called Model-Checking in the Loop Model-Based Testing (MCIL-MBT). In MCIL-MBT, we create a model that formalizes specifications of automotive operating systems and verifies the specifications via model-checking. Next, we conduct model-based testing with the verified model to ensure that a specific operating system implementation conforms to the model. These verification and testing stages are iterated over until no flaws are detected. Our method has already been introduced to an automotive system supplier and an operating system vendor. Through our approach, we successfully identified flaws that were not detected by conventional review and testing methods.

We study policy evaluation of offline contextual bandits subject to unobserved confounders. Sensitivity analysis methods are commonly used to estimate the policy value under the worst-case confounding over a given uncertainty set. However, existing work often resorts to some coarse relaxation of the uncertainty set for the sake of tractability, leading to overly conservative estimation of the policy value. In this paper, we propose a general estimator that provides a sharp lower bound of the policy value using convex programming. The generality of our estimator enables various extensions such as sensitivity analysis with f-divergence, model selection with cross validation and information criterion, and robust policy learning with the sharp lower bound. Furthermore, our estimation method can be reformulated as an empirical risk minimization problem thanks to the strong duality, which enables us to provide strong theoretical guarantees of the proposed estimator using techniques of the M-estimation., Comment: This is an extended version of the following work https://proceedings.mlr.press/v206/ishikawa23a.html. arXiv admin note: text overlap with arXiv:2302.13348

This paper proposes a new framework for learning a rule ensemble model that is both accurate and interpretable. A rule ensemble is an interpretable model based on the linear combination of weighted rules. In practice, we often face the trade-off between the accuracy and interpretability of rule ensembles. That is, a rule ensemble needs to include a sufficiently large number of weighted rules to maintain its accuracy, which harms its interpretability for human users. To avoid this trade-off and learn an interpretable rule ensemble without degrading accuracy, we introduce a new concept of interpretability, named local interpretability, which is evaluated by the total number of rules necessary to express individual predictions made by the model, rather than to express the model itself. Then, we propose a regularizer that promotes local interpretability and develop an efficient algorithm for learning a rule ensemble with the proposed regularizer by coordinate descent with local search. Experimental results demonstrated that our method learns rule ensembles that can explain individual predictions with fewer rules than the existing methods, including RuleFit, while maintaining comparable accuracy., Comment: 23 pages, 12 figures, to appear in the 2023 European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECMLPKDD 2023)

This paper tackles text-guided control of StyleGAN for editing garments in full-body human images. Existing StyleGAN-based methods suffer from handling the rich diversity of garments and body shapes and poses. We propose a framework for text-guided full-body human image synthesis via an attention-based latent code mapper, which enables more disentangled control of StyleGAN than existing mappers. Our latent code mapper adopts an attention mechanism that adaptively manipulates individual latent codes on different StyleGAN layers under text guidance. In addition, we introduce feature-space masking at inference time to avoid unwanted changes caused by text inputs. Our quantitative and qualitative evaluations reveal that our method can control generated images more faithfully to given texts than existing methods., Comment: VISIAPP 2024, project page: https://www.cgg.cs.tsukuba.ac.jp/~yoshikawa/pub/style_human_clip/

This paper proposes a new framework of algorithmic recourse (AR) that works even in the presence of missing values. AR aims to provide a recourse action for altering the undesired prediction result given by a classifier. Existing AR methods assume that we can access complete information on the features of an input instance. However, we often encounter missing values in a given instance (e.g., due to privacy concerns), and previous studies have not discussed such a practical situation. In this paper, we first empirically and theoretically show the risk that a naive approach with a single imputation technique fails to obtain good actions regarding their validity, cost, and features to be changed. To alleviate this risk, we formulate the task of obtaining a valid and low-cost action for a given incomplete instance by incorporating the idea of multiple imputation. Then, we provide some theoretical analyses of our task and propose a practical solution based on mixed-integer linear optimization. Experimental results demonstrated the efficacy of our method in the presence of missing values compared to the baselines., Comment: 30 pages, 15 figures

Representation learning has been increasing its impact on the research and practice of machine learning, since it enables to learn representations that can apply to various downstream tasks efficiently. However, recent works pay little attention to the fact that real-world datasets used during the stage of representation learning are commonly contaminated by noise, which can degrade the quality of learned representations. This paper tackles the problem to learn robust representations against noise in a raw dataset. To this end, inspired by recent works on denoising and the success of the cosine-similarity-based objective functions in representation learning, we propose the denoising Cosine-Similarity (dCS) loss. The dCS loss is a modified cosine-similarity loss and incorporates a denoising property, which is supported by both our theoretical and empirical findings. To make the dCS loss implementable, we also construct the estimators of the dCS loss with statistical guarantees. Finally, we empirically show the efficiency of the dCS loss over the baseline objective functions in vision and speech domains.

Contrastive learning is an efficient approach to self-supervised representation learning. Although recent studies have made progress in the theoretical understanding of contrastive learning, the investigation of how to characterize the clusters of the learned representations is still limited. In this paper, we aim to elucidate the characterization from theoretical perspectives. To this end, we consider a kernel-based contrastive learning framework termed Kernel Contrastive Learning (KCL), where kernel functions play an important role when applying our theoretical results to other frameworks. We introduce a formulation of the similarity structure of learned representations by utilizing a statistical dependency viewpoint. We investigate the theoretical properties of the kernel-based contrastive loss via this formulation. We first prove that the formulation characterizes the structure of representations learned with the kernel-based contrastive learning framework. We show a new upper bound of the classification error of a downstream task, which explains that our theory is consistent with the empirical success of contrastive learning. We also establish a generalization error bound of KCL. Finally, we show a guarantee for the generalization ability of KCL to the downstream classification task via a surrogate bound.