Adversarial training (AT) is widely recognized as one of the most effective methods for improving the robustness of deep learning models. However, AT suffers from a fundamental trade-off between robustness and generalization, which has motivated various mitigation strategies. Among them, margin-based AT approaches employ loss reweighting, reflecting the idea that more critical examples should contribute larger gradients. Yet, these methods are limited by their exclusive focus on gradient magnitude. In this work, we identify that prior approaches overlook the role of gradient direction, and we provide both theoretical and empirical evidence to support this claim. We argue that both the magnitude and direction of gradients should be considered in adversarial training, and propose a novel label design framework, ADA-Lab ( ADversary’s Adversary for Label adjustment ), which incorporates both aspects to refine supervision for low-margin examples. Specifically, we introduce the concept of the adversary’s adversary to explicitly encode directional information aligned with gradient descent. Our theoretical analysis shows that labels designed using this concept better approximate the true label distribution, especially for low-margin examples (i.e., more important examples). Furthermore, by estimating example importance based on the distance to the decision boundary, our method adaptively controls the degree of label interpolation. Our key novelty lies in introducing direction-aware label refinement based on the adversary’s adversary, a concept that explicitly leverages the gradient descent direction of adversarial inputs to correct label mismatch. This unified design integrates gradient magnitude-based importance weighting and label distribution correction, resulting in improved robustness and generalization, as demonstrated by extensive theoretical and empirical results. • We propose a new perspective on adversarial training by introducing direction-aware label refinement based on the adversary’s adversary concept, which has not been explored in prior margin-based or label smoothing methods. • Our framework unifies the benefits of gradient magnitude-based importance weighting and label distribution correction, offering a principled and scalable approach to improving adversarial robustness. • We provide both theoretical and empirical evidence that reducing margin variance and distribution mismatch leads to a tighter bound on natural and robust risk.

This work presents a ring oscillator (RO)-based low-jitter injection-locked clock and data recovery (ILCDR) with a pattern-dependent pulse filtering (PDPF) technique. The conventional ILCDR has a drawback that data jitter is transferred to the recovered clock. To reduce jitter, the PDPF technique is employed to filter out the injection pulses occurring in data patterns that cause high data-dependent jitter (DDJ). Adopting the PDPF technique with an injection timing control loop, the ILCDR optimizes injection timing and maximizes timing margin. Fabricated in a 28-nm CMOS technology, the proposed ILCDR occupies an active area of 0.03 mm2and consumes 13.6 mW at 10 Gb/s. The measured jitter tolerance (JTOL) is 1 UIppat 35 MHz with a bit error rate (BER) of$10^{-12}$.

Deep learning-based computer vision systems adopt complex and large architectures to improve performance, yet they face challenges in deployment on resource-constrained mobile and edge devices. To address this issue, model compression techniques such as pruning, quantization, and matrix factorization have been proposed; however, these compressed models are often highly vulnerable to adversarial attacks. We introduce the Efficient Ensemble Defense (EED) technique, which diversifies the compression of a single base model based on different pruning importance scores and enhances ensemble diversity to achieve high adversarial robustness and resource efficiency. EED dynamically determines the number of necessary sub-models during the inference stage, minimizing unnecessary computations while maintaining high robustness. On the CIFAR- 10 and SVHN datasets, EED demonstrated state-of-the-art robustness performance compared to existing adversarial pruning techniques, along with an inference speed improvement of up to 1.86 times. This proves that EED is a powerful defense solution in resource-constrained environments.

This paper presents a referenceless baud-rate clock and data recovery (CDR) circuit with an integrated pattern decoding technique for fast frequency acquisition. The proposed CDR achieves baud-rate frequency acquisition without edge sampling clock phase by utilizing an integrator circuit. The proposed integrated pattern decoding technique divides 6-bit sequential patterns into 5 pattern groups. The patterns of each pattern group are detected from the specific frequency offsets. By applying larger weights to the pattern groups including the patterns detected at the higher frequency offset, the proposed referenceless CDR achieves fast frequency acquisition. Fabricated in a 28-nm CMOS technology, the CDR prototype occupies 0.039 mm2and consumes 30.77 mW at 30 Gb/s. From various initial clock frequency, the proposed CDR achieves a capture range from 16 Gb/s to 30 Gb/s. Thanks to the fast frequency acquisition ability of the proposed integrated pattern decoding technique, the worst frequency acquisition time at the data rate of 30 Gb/s is$3.7~\mu $s. The CDR achieves a bit error rate (BER) of less than$10^{-12}$and an energy efficiency of 1.03 pJ/b.

In this paper, we propose DIRE (Domain-Invariant Representation Learning for Expression), a novel approach to enhance the generalizability of facial expression recognition (FER) models in unseen domains. Traditional FER models often struggle with distribution shifts between training and test datasets, leading to significant performance drops. Based on the concept of Single-Source Domain Generalization, we introduce a novel domain augmentation technique that applies pixel-level and feature-level perturbations to domain-variant regions while preserving semantic consistency. Additionally, we incorporate semantic alignment regularization and domain information minimization loss so that domain-invariant features effectively represent facial expressions. Extensive experiments on multiple FER datasets demonstrate that our method significantly improves generalization across diverse target domains, even when trained on a single source domain. The proposed DIRE approach offers a robust solution to real-world FER tasks, where unseen domain generalizability is crucial.

Beyond the in-the-lab environment, deep-learning-based facial expression recognition (FER) models that provide reliable performance on wild datasets are gradually becoming applied to the real world. However, the fact that neural networks are inherently vulnerable to digital attacks (e.g., adversarial examples) and their performance is not exposed to external threats reduces the applicability of FER technology. So, we design a so-called test-time attack scenario in which FER models are deceived by superimposing imperceptible perturbation(s) on test images. This scenario, which targets the testing phase in which model weakness is revealed, clearly shows how vulnerable FER models are to external attacks. As a remedy against this attack, we propose a novel method called FAAT, which adversarially trains the model by paying attention to core region(s) of face. FAAT aims to improve model robustness so that the model can be generalized to unseen perturbation(s) while focusing on facial expression-related areas. For example, FAAT’s robustness against PGD attack with a performance improvement of up to 18% is encouraging. Also, various benchmarking results based on our attack scenario analyze the fidelity of prior arts and will promote the development direction of future models.

A change from a non-return-to-zero (NRZ) signaling to a four-level pulse amplitude modulation (PAM-4) signaling causes various challenges in clock and data recovery (CDR) designs as well as analog-front-end (AFE) designs. A PAM-4 CDR with a 2x-oversampling phase detector (PD) has an issue of increased pattern-dependent jitter due to asymmetric transitions. This work investigates a similar problem in a baud-rate CDR by analyzing the PD characteristics. In the PAM-4 baud-rate sampling, the transitions are classified into two types: full-swing transitions and non-full-swing transitions. Since utilizing the non-full-swing transitions can affect the jitter tracking ability, careful consideration of decisions using these transitions is necessary to optimize the jitter performance. To address this issue, we propose an asymmetric-weighted PD that minimizes pattern-dependent jitter and maximizes a transition density by utilizing both the full-swing transitions and the non-full-swing transitions. Using a pseudo-linear analysis, the proposed PD achieves improved jitter performance compared to the conventional PD. Fabricated in 28-nm CMOS process, a prototype PAM-4 receiver with the proposed CDR is demonstrated at 40 Gb/s. The CDR achieves a bit error rate (BER) less than 10$^{-9}$and an energy efficiency of 1.65 pJ/b.