Pei-Yuan Wu

Hongxin Lin, Yunwei Chiu, Peiyuan Wu

The graph convolutional networks (GCNs) have been applied to model the physically connected and non-local relations among human joints for 3D human pose estimation (HPE). In addition, the purely Transformer-based models recently show promising results in video-based 3D HPE. However, the single-frame method still needs to model the physically connected relations among joints because the feature representations transformed only by global relations via the Transformer neglect information on the human skeleton. To deal with this problem, we propose a novel method in which the Transformer encoder and GCN blocks are alternately stacked, namely AMPose, to combine the global and physically connected relations among joints towards HPE. In the AMPose, the Transformer encoder is applied to connect each joint with all the other joints, while GCNs are applied to capture information on physically connected relations. The effectiveness of our proposed method is evaluated on the Human3.6M dataset. Our model also shows better generalization ability by testing on the MPI-INF-3DHP dataset. Code can be retrieved at https://github.com/erikervalid/AMPose.

Sing-Yuan Yeh, Fu-Chieh Chang, Chang-Wei Yueh et al.

Modern reinforcement learning (RL) often faces an enormous state-action space. Existing analytical results are typically for settings with a small number of state-actions, or simple models such as linearly modeled Q-functions. To derive statistically efficient RL policies handling large state-action spaces, with more general Q-functions, some recent works have considered nonlinear function approximation using kernel ridge regression. In this work, we derive sample complexities for kernel based Q-learning when a generative model exists. We propose a nonparametric Q-learning algorithm which finds an $ε$-optimal policy in an arbitrarily large scale discounted MDP. The sample complexity of the proposed algorithm is order optimal with respect to $ε$ and the complexity of the kernel (in terms of its information gain). To the best of our knowledge, this is the first result showing a finite sample complexity under such a general model.

Yu-Ting Lee, Fu-Chieh Chang, Hui-Ying Shih et al.

Intrinsic self-correction refers to the phenomenon where a language model refines its own outputs purely through prompting, without external feedback or parameter updates. While this approach improves performance across diverse tasks, its internal mechanism remains poorly understood. We analyze intrinsic self-correction from a representation-level perspective. We formalize and introduce the notion of a prompt-induced shift, which is the change in hidden representations caused by a self-correction prompt. Across 5 open-source LLMs, prompt-induced shifts in text detoxification and text toxification align with latent directions constructed from contrastive pairs. In detoxification, the shifts align with the non-toxic direction; in toxification, they align with the toxic direction. These results suggest that intrinsic self-correction functions as representation steering along interpretable latent directions, beyond what standard metrics such as task scores or model confidence capture. Our analysis offers an interpretability-based account of intrinsic self-correction and contributes to a more systematic understanding of LLM prompting.

Yen-Ru Lai, Fu-Chieh Chang, Pei-Yuan Wu

Offline reinforcement learning (RL) learns policies from a fixed dataset, but often requires large amounts of data. The challenge arises when labeled datasets are expensive, especially when rewards have to be provided by human labelers for large datasets. In contrast, unlabelled data tends to be less expensive. This situation highlights the importance of finding effective ways to use unlabelled data in offline RL, especially when labelled data is limited or expensive to obtain. In this paper, we present the algorithm to utilize the unlabeled data in the offline RL method with kernel function approximation and give the theoretical guarantee. We present various eigenvalue decay conditions of $\mathcal{H}_k$ which determine the complexity of the algorithm. In summary, our work provides a promising approach for exploiting the advantages offered by unlabeled data in offline RL, whilst maintaining theoretical assurances.

Hung Nguyen, Peiyuan Wu, Morris Chang

One of the most challenging issues in federated learning is that the data is often not independent and identically distributed (nonIID). Clients are expected to contribute the same type of data and drawn from one global distribution. However, data are often collected in different ways from different resources. Thus, the data distributions among clients might be different from the underlying global distribution. This creates a weight divergence issue and reduces federated learning performance. This work focuses on improving federated learning performance for skewed data distribution across clients. The main idea is to adjust the client distribution closer to the global distribution using sample weights. Thus, the machine learning model converges faster with higher accuracy. We start from the fundamental concept of empirical risk minimization and theoretically derive a solution for adjusting the distribution skewness using sample weights. To determine sample weights, we implicitly exchange density information by leveraging a neural network-based density estimation model, MADE. The clients data distribution can then be adjusted without exposing their raw data. Our experiment results on three real-world datasets show that the proposed method not only improves federated learning accuracy but also significantly reduces communication costs compared to the other experimental methods.

Chia-Wen Huang, Haw Hwai, Chien-Chang Lee et al.

Timely and accurate diagnosis of appendicitis is critical in clinical settings to prevent serious complications. While CT imaging remains the standard diagnostic tool, the growing number of cases can overwhelm radiologists, potentially causing delays. In this paper, we propose a deep learning model that leverages 3D CT scans for appendicitis classification, incorporating Slice Attention mechanisms guided by external 2D datasets to enhance small lesion detection. Additionally, we introduce a hierarchical classification framework using pre-trained 2D models to differentiate between simple and complicated appendicitis. Our approach improves AUC by 3% for appendicitis and 5.9% for complicated appendicitis, offering a more efficient and reliable diagnostic solution compared to previous work.

Yu Wang, Fu-Chieh Chang, Pei-Yuan Wu

We study the robustness of Transformer language models under semantic out-of-distribution (OOD) shifts, where training and test data lie in disjoint latent spaces. Using Wasserstein-1 distance and Gevrey-class smoothness, we derive sub-exponential upper bounds on prediction error. Our theoretical framework explains how smoothness governs generalization under distributional drift. We validate these findings through controlled experiments on arithmetic and Chain-of-Thought tasks with latent permutations and scalings. Results show empirical degradation aligns with our bounds, highlighting the geometric and functional principles underlying OOD generalization in Transformers.

Jia-Hau Bai, Chi-Ting Liu, Yu Wang et al.

This study uses CAPM (Convex Adversarial Polytope for Maxpool-based CNN) to improve the verified bound for general purpose maxpool-based convolutional neural networks (CNNs) under bounded norm adversarial perturbations. The maxpool function is decomposed as a series of ReLU functions to extend the convex relaxation technique to maxpool functions, by which the verified bound can be efficiently computed through a dual network. The experimental results demonstrate that this technique allows the state-of-the-art verification precision for maxpool-based CNNs and involves a much lower computational cost than current verification methods, such as DeepZ, DeepPoly and PRIMA. This method is also applicable to large-scale CNNs, which previous studies show to be often computationally prohibitively expensive. Under certain circumstances, CAPM is 40-times, 20-times or twice as fast and give a significantly higher verification bound (CAPM 98% vs. PRIMA 76%/DeepPoly 73%/DeepZ 8%) as compared to PRIMA/DeepPoly/DeepZ. Furthermore, we additionally present the time complexity of our algorithm as $O(W^2NK)$, where $W$ is the maximum width of the neural network, $N$ is the number of neurons, and $K$ is the size of the maxpool layer's kernel.

Codruta O. Ancuti, Cosmin Ancuti, Florin-Alexandru Vasluianu et al.

This paper reviews the NTIRE 2020 Challenge on NonHomogeneous Dehazing of images (restoration of rich details in hazy image). We focus on the proposed solutions and their results evaluated on NH-Haze, a novel dataset consisting of 55 pairs of real haze free and nonhomogeneous hazy images recorded outdoor. NH-Haze is the first realistic nonhomogeneous haze dataset that provides ground truth images. The nonhomogeneous haze has been produced using a professional haze generator that imitates the real conditions of haze scenes. 168 participants registered in the challenge and 27 teams competed in the final testing phase. The proposed solutions gauge the state-of-the-art in image dehazing.

Hung Nguyen, Di Zhuang, Pei-Yuan Wu et al.

Protecting sensitive information against data exploiting attacks is an emerging research area in data mining. Over the past, several different methods have been introduced to protect individual privacy from such attacks while maximizing data-utility of the application. However, these existing techniques are not sufficient to effectively protect data owner privacy, especially in the scenarios that utilize visualizable data (e.g. images, videos) or the applications that require heavy computations for implementation. To address these problems, we propose a new dimension reduction-based method for privacy preservation. Our method generates dimension-reduced data for performing machine learning tasks and prevents a strong adversary from reconstructing the original data. We first introduce a theoretical approach to evaluate dimension reduction-based privacy preserving mechanisms, then propose a non-linear dimension reduction framework motivated by state-of-the-art neural network structures for privacy preservation. We conducted experiments over three different face image datasets (AT&T, YaleB, and CelebA), and the results show that when the number of dimensions is reduced to seven, we can achieve the accuracies of 79%, 80%, and 73% respectively and the reconstructed images are not recognizable to naked human eyes.