Hsuan-Tien Lin

Wei-Chao Cheng, Tan-Ha Mai, Hsuan-Tien Lin

Given imbalanced data, it is hard to train a good classifier using deep learning because of the poor generalization of minority classes. Traditionally, the well-known synthetic minority oversampling technique (SMOTE) for data augmentation, a data mining approach for imbalanced learning, has been used to improve this generalization. However, it is unclear whether SMOTE also benefits deep learning. In this work, we study why the original SMOTE is insufficient for deep learning, and enhance SMOTE using soft labels. Connecting the resulting soft SMOTE with Mixup, a modern data augmentation technique, leads to a unified framework that puts traditional and modern data augmentation techniques under the same umbrella. A careful study within this framework shows that Mixup improves generalization by implicitly achieving uneven margins between majority and minority classes. We then propose a novel margin-aware Mixup technique that more explicitly achieves uneven margins. Extensive experimental results demonstrate that our proposed technique yields state-of-the-art performance on deep imbalanced classification while achieving superior performance on extremely imbalanced data. The code is open-sourced in our developed package https://github.com/ntucllab/imbalanced-DL to foster future research in this direction.

Andrew Bai, Cho-Jui Hsieh, Wendy Kan et al.

Generative adversarial network (GAN) continues to be a popular research direction due to its high generation quality. It is observed that many state-of-the-art GANs generate samples that are more similar to the training set than a holdout testing set from the same distribution, hinting some training samples are implicitly memorized in these models. This memorization behavior is unfavorable in many applications that demand the generated samples to be sufficiently distinct from known samples. Nevertheless, it is unclear whether it is possible to reduce memorization without compromising the generation quality. In this paper, we propose memorization rejection, a training scheme that rejects generated samples that are near-duplicates of training samples during training. Our scheme is simple, generic and can be directly applied to any GAN architecture. Experiments on multiple datasets and GAN models validate that memorization rejection effectively reduces training sample memorization, and in many cases does not sacrifice the generation quality. Code to reproduce the experiment results can be found at $\texttt{https://github.com/jybai/MRGAN}$.

Yu-Chu Yu, Hsuan-Tien Lin

Semi-Supervised Domain Adaptation (SSDA) involves learning to classify unseen target data with a few labeled and lots of unlabeled target data, along with many labeled source data from a related domain. Current SSDA approaches usually aim at aligning the target data to the labeled source data with feature space mapping and pseudo-label assignments. Nevertheless, such a source-oriented model can sometimes align the target data to source data of the wrong classes, degrading the classification performance. This paper presents a novel source-adaptive paradigm that adapts the source data to match the target data. Our key idea is to view the source data as a noisily-labeled version of the ideal target data. Then, we propose an SSDA model that cleans up the label noise dynamically with the help of a robust cleaner component designed from the target perspective. Since the paradigm is very different from the core ideas behind existing SSDA approaches, our proposed model can be easily coupled with them to improve their performance. Empirical results on two state-of-the-art SSDA approaches demonstrate that the proposed model effectively cleans up the noise within the source labels and exhibits superior performance over those approaches across benchmark datasets. Our code is available at https://github.com/chu0802/SLA .

Po-Yi Lu, Yi-Jie Cheng, Chun-Liang Li et al.

Active Learning (AL) addresses the crucial challenge of enabling machines to efficiently gather labeled examples through strategic queries. Among the many AL strategies, Uncertainty Sampling (US) stands out as one of the most widely adopted. US queries the example(s) that the current model finds uncertain, proving to be both straightforward and effective. Despite claims in the literature suggesting superior alternatives to US, community-wide acceptance remains elusive. In fact, existing benchmarks for tabular datasets present conflicting conclusions on the continued competitiveness of US. In this study, we review the literature on AL strategies in the last decade and build the most comprehensive open-source AL benchmark to date to understand the relative merits of different AL strategies. The benchmark surpasses existing ones by encompassing a broader coverage of strategies, models, and data. Through our investigation of the conflicting conclusions in existing tabular AL benchmarks by evaluation under broad AL experimental settings, we uncover fresh insights into the often-overlooked issue of using machine learning models--**model compatibility** in the context of US. Specifically, we notice that adopting the different models for the querying unlabeled examples and learning tasks would degrade US's effectiveness. Notably, our findings affirm that US maintains a competitive edge over other strategies when paired with compatible models. These findings have practical implications and provide a concrete recipe for AL practitioners, empowering them to make informed decisions when working with tabular classifications with limited labeled data. The code for this project is available on https://github.com/ariapoy/active-learning-benchmark.

Xiwei Xuan, Ziquan Deng, Hsuan-Tien Lin et al.

Recent studies highlight that deep learning models often learn spurious features mistakenly linked to labels, compromising their reliability in real-world scenarios where such correlations do not hold. Despite the increasing research effort, existing solutions often face two main challenges: they either demand substantial annotations of spurious attributes, or they yield less competitive outcomes with expensive training when additional annotations are absent. In this paper, we introduce SLIM, a cost-effective and performance-targeted approach to reducing spurious correlations in deep learning. Our method leverages a human-in-the-loop protocol featuring a novel attention labeling mechanism with a constructed attention representation space. SLIM significantly reduces the need for exhaustive additional labeling, requiring human input for fewer than 3% of instances. By prioritizing data quality over complicated training strategies, SLIM curates a smaller yet more feature-balanced data subset, fostering the development of spuriousness-robust models. Experimental validations across key benchmarks demonstrate that SLIM competes with or exceeds the performance of leading methods while significantly reducing costs. The SLIM framework thus presents a promising path for developing reliable models more efficiently. Our code is available in https://github.com/xiweix/SLIM.git/.

Oscar Chew, Po-Yi Lu, Jayden Lin et al.

Text-to-image diffusion models have been widely adopted in real-world applications due to their ability to generate realistic images from textual descriptions. However, recent studies have shown that these methods are vulnerable to backdoor attacks. Despite the significant threat posed by backdoor attacks on text-to-image diffusion models, countermeasures remain under-explored. In this paper, we address this research gap by demonstrating that state-of-the-art backdoor attacks against text-to-image diffusion models can be effectively mitigated by a surprisingly simple defense strategy - textual perturbation. Experiments show that textual perturbations are effective in defending against state-of-the-art backdoor attacks with minimal sacrifice to generation quality. We analyze the efficacy of textual perturbation from two angles: text embedding space and cross-attention maps. They further explain how backdoor attacks have compromised text-to-image diffusion models, providing insights for studying future attack and defense strategies. Our code is available at https://github.com/oscarchew/t2i-backdoor-defense.

Xiwei Xuan, Ziquan Deng, Hsuan-Tien Lin et al.

Researchers have proposed various methods for visually interpreting the Convolutional Neural Network (CNN) via saliency maps, which include Class-Activation-Map (CAM) based approaches as a leading family. However, in terms of the internal design logic, existing CAM-based approaches often overlook the causal perspective that answers the core "why" question to help humans understand the explanation. Additionally, current CNN explanations lack the consideration of both necessity and sufficiency, two complementary sides of a desirable explanation. This paper presents a causality-driven framework, SUNY, designed to rationalize the explanations toward better human understanding. Using the CNN model's input features or internal filters as hypothetical causes, SUNY generates explanations by bi-directional quantifications on both the necessary and sufficient perspectives. Extensive evaluations justify that SUNY not only produces more informative and convincing explanations from the angles of necessity and sufficiency, but also achieves performances competitive to other approaches across different CNN architectures over large-scale datasets, including ILSVRC2012 and CUB-200-2011.

Vo Nguyen Le Duy, Hsuan-Tien Lin, Ichiro Takeuchi

We propose a novel statistical method for testing the results of anomaly detection (AD) under domain adaptation (DA), which we call CAD-DA -- controllable AD under DA. The distinct advantage of the CAD-DA lies in its ability to control the probability of misidentifying anomalies under a pre-specified level $α$ (e.g., 0.05). The challenge within this DA setting is the necessity to account for the influence of DA to ensure the validity of the inference results. Our solution to this challenge leverages the concept of conditional Selective Inference to handle the impact of DA. To our knowledge, this is the first work capable of conducting a valid statistical inference within the context of DA. We evaluate the performance of the CAD-DA method on both synthetic and real-world datasets.

Wei-I Lin, Hsuan-Tien Lin

Complementary-Label Learning (CLL) is a weakly-supervised learning problem that aims to learn a multi-class classifier from only complementary labels, which indicate a class to which an instance does not belong. Existing approaches mainly adopt the paradigm of reduction to ordinary classification, which applies specific transformations and surrogate losses to connect CLL back to ordinary classification. Those approaches, however, face several limitations, such as the tendency to overfit or be hooked on deep models. In this paper, we sidestep those limitations with a novel perspective--reduction to probability estimates of complementary classes. We prove that accurate probability estimates of complementary labels lead to good classifiers through a simple decoding step. The proof establishes a reduction framework from CLL to probability estimates. The framework offers explanations of several key CLL approaches as its special cases and allows us to design an improved algorithm that is more robust in noisy environments. The framework also suggests a validation procedure based on the quality of probability estimates, leading to an alternative way to validate models with only complementary labels. The flexible framework opens a wide range of unexplored opportunities in using deep and non-deep models for probability estimates to solve the CLL problem. Empirical experiments further verified the framework's efficacy and robustness in various settings.

Paul Kuo-Ming Huang, Si-An Chen, Hsuan-Tien Lin

Score-based generative models (SGMs) are a popular family of deep generative models that achieve leading image generation quality. Early studies extend SGMs to tackle class-conditional generation by coupling an unconditional SGM with the guidance of a trained classifier. Nevertheless, such classifier-guided SGMs do not always achieve accurate conditional generation, especially when trained with fewer labeled data. We argue that the problem is rooted in the classifier's tendency to overfit without coordinating with the underlying unconditional distribution. To make the classifier respect the unconditional distribution, we propose improving classifier-guided SGMs by letting the classifier regularize itself. The key idea of our proposed method is to use principles from energy-based models to convert the classifier into another view of the unconditional SGM. Existing losses for unconditional SGMs can then be leveraged to achieve regularization by calibrating the classifier's internal unconditional scores. The regularization scheme can be applied to not only the labeled data but also unlabeled ones to further improve the classifier. Across various percentages of fewer labeled data, empirical results show that the proposed approach significantly enhances conditional generation quality. The enhancements confirm the potential of the proposed self-calibration technique for generative modeling with limited labeled data.

Tan-Ha Mai, Chao-Kai Chiang, Han-Hwa Shih et al.

Complementary-label learning (CLL) is a weakly supervised paradigm where instances are labeled with classes they do not belong to. Despite a decade of research, CLL methods remain competitive mainly on 10-class classification, with scaling to large label spaces continuing to be an enduring bottleneck. This limitation stems from the common assumption of uniform label generation in traditional methods, which fatally dilutes the learning signal in many-class settings. In this paper, we demonstrate that this long-standing barrier can be overcome by deliberately designing a biased (non-uniform) generation process that restricts complementary labels to a subset of classes. This finding motivates us to propose Bias-Induced Constrained Labeling (BICL), a principled framework spanning data collection to training that leverages this bias. BICL enables effective learning on CIFAR-100 and TinyImageNet-200, achieving more than sevenfold accuracy improvements over traditional methods. Our findings establish a new trajectory for making CLL feasible for many classes in real-world applications.

Pin-Han Huang, Shang-Tse Chen, Hsuan-Tien Lin

Incorporating diffusion-generated synthetic data into adversarial training (AT) has been shown to substantially improve the training of robust image classifiers. In this work, we extend the role of diffusion models beyond merely generating synthetic data, examining whether their internal representations, which encode meaningful features of the data, can provide additional benefits for robust classifier training. Through systematic experiments, we show that diffusion models offer representations that are both diverse and partially robust, and that explicitly incorporating diffusion representations as an auxiliary learning signal during AT consistently improves robustness across settings. Furthermore, our representation analysis indicates that incorporating diffusion models into AT encourages more disentangled features, while diffusion representations and diffusion-generated synthetic data play complementary roles in shaping representations. Experiments on CIFAR-10, CIFAR-100, and ImageNet validate these findings, demonstrating the effectiveness of jointly leveraging diffusion representations and synthetic data within AT.

Oscar Chew, Hsuan-Tien Lin, Kai-Wei Chang et al.

Recent research has revealed that machine learning models have a tendency to leverage spurious correlations that exist in the training set but may not hold true in general circumstances. For instance, a sentiment classifier may erroneously learn that the token "performances" is commonly associated with positive movie reviews. Relying on these spurious correlations degrades the classifiers performance when it deploys on out-of-distribution data. In this paper, we examine the implications of spurious correlations through a novel perspective called neighborhood analysis. The analysis uncovers how spurious correlations lead unrelated words to erroneously cluster together in the embedding space. Driven by the analysis, we design a metric to detect spurious tokens and also propose a family of regularization methods, NFL (doN't Forget your Language) to mitigate spurious correlations in text classification. Experiments show that NFL can effectively prevent erroneous clusters and significantly improve the robustness of classifiers without auxiliary data. The code is publicly available at https://github.com/oscarchew/doNt-Forget-your-Language.

Hsiu-Hsuan Wang, Tan-Ha Mai, Nai-Xuan Ye et al.

Complementary-label learning (CLL) is a weakly-supervised learning paradigm that aims to train a multi-class classifier using only complementary labels, which indicate classes to which an instance does not belong. Despite numerous algorithmic proposals for CLL, their practical applicability remains unverified for two reasons. Firstly, these algorithms often rely on assumptions about the generation of complementary labels, and it is not clear how far the assumptions are from reality. Secondly, their evaluation has been limited to synthetically labeled datasets. To gain insights into the real-world performance of CLL algorithms, we developed a protocol to collect complementary labels from human annotators. Our efforts resulted in the creation of four datasets: CLCIFAR10, CLCIFAR20, CLMicroImageNet10, and CLMicroImageNet20, derived from well-known classification datasets CIFAR10, CIFAR100, and TinyImageNet200. These datasets represent the very first real-world CLL datasets, namely CLImage, which are publicly available at: https://github.com/ntucllab/CLImage\_Dataset. Through extensive benchmark experiments, we discovered a notable decrease in performance when transitioning from synthetically labeled datasets to real-world datasets. We investigated the key factors contributing to the decrease with a thorough dataset-level ablation study. Our analyses highlight annotation noise as the most influential factor in the real-world datasets. In addition, we discover that the biased-nature of human-annotated complementary labels and the difficulty to validate with only complementary labels are two outstanding barriers to practical CLL. These findings suggest that the community focus more research efforts on developing CLL algorithms and validation schemes that are robust to noisy and biased complementary-label distributions.

Si-An Chen, Chun-Liang Li, Hsuan-Tien Lin

Conditional Generative Adversarial Networks (cGANs) are implicit generative models which allow to sample from class-conditional distributions. Existing cGANs are based on a wide range of different discriminator designs and training objectives. One popular design in earlier works is to include a classifier during training with the assumption that good classifiers can help eliminate samples generated with wrong classes. Nevertheless, including classifiers in cGANs often comes with a side effect of only generating easy-to-classify samples. Recently, some representative cGANs avoid the shortcoming and reach state-of-the-art performance without having classifiers. Somehow it remains unanswered whether the classifiers can be resurrected to design better cGANs. In this work, we demonstrate that classifiers can be properly leveraged to improve cGANs. We start by using the decomposition of the joint probability distribution to connect the goals of cGANs and classification as a unified framework. The framework, along with a classic energy model to parameterize distributions, justifies the use of classifiers for cGANs in a principled manner. It explains several popular cGAN variants, such as ACGAN, ProjGAN, and ContraGAN, as special cases with different levels of approximations, which provides a unified view and brings new insights to understanding cGANs. Experimental results demonstrate that the design inspired by the proposed framework outperforms state-of-the-art cGANs on multiple benchmark datasets, especially on the most challenging ImageNet. The code is available at https://github.com/sian-chen/PyTorch-ECGAN.

Ashesh, Chu-Song Chen, Hsuan-Tien Lin

Gaze estimation involves predicting where the person is looking at within an image or video. Technically, the gaze information can be inferred from two different magnification levels: face orientation and eye orientation. The inference is not always feasible for gaze estimation in the wild, given the lack of clear eye patches in conditions like extreme left/right gazes or occlusions. In this work, we design a model that mimics humans' ability to estimate the gaze by aggregating from focused looks, each at a different magnification level of the face area. The model avoids the need to extract clear eye patches and at the same time addresses another important issue of face-scale variation for gaze estimation in the wild. We further extend the model to handle the challenging task of 360-degree gaze estimation by encoding the backward gazes in the polar representation along with a robust averaging scheme. Experiment results on the ETH-XGaze dataset, which does not contain scale-varying faces, demonstrate the model's effectiveness to assimilate information from multiple scales. For other benchmark datasets with many scale-varying faces (Gaze360 and RT-GENE), the proposed model achieves state-of-the-art performance for gaze estimation when using either images or videos. Our code and pretrained models can be accessed at https://github.com/ashesh-0/MultiZoomGaze.

Yao-Yuan Yang, Shao-Chuan Lee, Yu-An Chung et al.

libact is a Python package designed to make active learning easier for general users. The package not only implements several popular active learning strategies, but also features the active-learning-by-learning meta-algorithm that assists the users to automatically select the best strategy on the fly. Furthermore, the package provides a unified interface for implementing more strategies, models and application-specific labelers. The package is open-source on Github, and can be easily installed from Python Package Index repository.

Oscar Chew, Po-Yi Lu, Jayden Lin et al.

Recent studies show that text to image (T2I) diffusion models are vulnerable to backdoor attacks, where a trigger in the input prompt can steer generation toward harmful or unintended content. Beyond the trigger token itself, backdoor effects can spread to neighboring tokens in the text embedding space. To address this, we introduce PEPPER (PErcePtion Guided PERturbation), a backdoor defense that rewrites the caption into a semantically distant yet visually similar caption while adding unobstructive elements. With this rewriting strategy, PEPPER disrupt the trigger embedded in the input prompt, dilute the influence of trigger tokens and thereby achieve enhanced robustness. Experiments show that PEPPER is particularly effective against text encoder based attacks, substantially reducing attack success while preserving generation quality. Beyond this, PEPPER can be paired with any existing defenses yielding consistently stronger and generalizable robustness than any standalone method. Our code will be released on Github.

Nai-Xuan Ye, Tan-Ha Mai, Hsiu-Hsuan Wang et al.

Complementary-label learning (CLL) is a weakly supervised learning paradigm for multiclass classification, where only complementary labels -- indicating classes an instance does not belong to -- are provided to the learning algorithm. Despite CLL's increasing popularity, previous studies highlight two main challenges: (1) inconsistent results arising from varied assumptions on complementary label generation, and (2) high barriers to entry due to the lack of a standardized evaluation platform across datasets and algorithms. To address these challenges, we introduce \texttt{libcll}, an extensible Python toolkit for CLL research. \texttt{libcll} provides a universal interface that supports a wide range of generation assumptions, both synthetic and real-world datasets, and key CLL algorithms. The toolkit is designed to mitigate inconsistencies and streamline the research process, with easy installation, comprehensive usage guides, and quickstart tutorials that facilitate efficient adoption and implementation of CLL techniques. Extensive ablation studies conducted with \texttt{libcll} demonstrate its utility in generating valuable insights to advance future CLL research.

An-Sheng Lee, Yu-Wen Pao, Hsuan-Tien Lin et al.

Pre-training foundation models has become the de-facto procedure for deep learning approaches, yet its application remains limited in the geological studies, where in needs of the model transferability to break the shackle of data scarcity. Here we target on the X-ray fluorescence (XRF) scanning data, a standard high-resolution measurement in extensive scientific drilling projects. We propose a scalable self-supervised learner, masked autoencoders on XRF spectra (MAX), to pre-train a foundation model covering geological records from multiple regions of the Pacific and Southern Ocean. In pre-training, we find that masking a high proportion of the input spectrum (50\%) yields a nontrivial and meaningful self-supervisory task. For downstream tasks, we select the quantification of XRF spectra into two costly geochemical measurements, CaCO$_3$ and total organic carbon, due to their importance in understanding the paleo-oceanic carbon system. Our results show that MAX, requiring only one-third of the data, outperforms models without pre-training in terms of quantification accuracy. Additionally, the model's generalizability improves by more than 60\% in zero-shot tests on new materials, with explainability further ensuring its robustness. Thus, our approach offers a promising pathway to overcome data scarcity in geological discovery by leveraging the self-supervised foundation model and fast-acquired XRF scanning data.

Hung-Chieh Fang, Po-Yi Lu, Hsuan-Tien Lin

Universal Domain Adaptation (UniDA) addresses unsupervised domain adaptation where target classes may differ arbitrarily from source ones, except for a shared subset. A widely used approach, partial domain matching (PDM), aligns only shared classes but struggles in extreme cases where many source classes are absent in the target domain, underperforming the most naive baseline that trains on only source data. In this work, we identify that the failure of PDM for extreme UniDA stems from dimensional collapse (DC) in target representations. To address target DC, we propose to use the de-collapse techniques in self-supervised learning on the unlabeled target data to preserve the intrinsic structure of the learned representations. Our experimental results confirm that SSL consistently advances PDM and delivers new state-of-the-art results across a broader benchmark of UniDA scenarios with different portions of shared classes, representing a crucial step toward truly comprehensive UniDA. Project page: https://dc-unida.github.io/

Tan-Ha Mai, Hsuan-Tien Lin

In this paper, we investigate the challenges of complementary-label learning (CLL), a specialized form of weakly-supervised learning (WSL) where models are trained with labels indicating classes to which instances do not belong, rather than standard ordinary labels. This alternative supervision is appealing because collecting complementary labels is generally cheaper and less labor-intensive. Although most existing research in CLL emphasizes the development of novel loss functions, the potential of data augmentation in this domain remains largely underexplored. In this work, we uncover that the widely-used Mixup data augmentation technique is ineffective when directly applied to CLL. Through in-depth analysis, we identify that the complementary-label noise generated by Mixup negatively impacts the performance of CLL models. We then propose an improved technique called Intra-Cluster Mixup (ICM), which only synthesizes augmented data from nearby examples, to mitigate the noise effect. ICM carries the benefits of encouraging complementary label sharing of nearby examples, and leads to substantial performance improvements across synthetic and real-world labeled datasets. In particular, our wide spectrum of experimental results on both balanced and imbalanced CLL settings justifies the potential of ICM in allying with state-of-the-art CLL algorithms, achieving significant accuracy increases of 30% and 10% on MNIST and CIFAR datasets, respectively.

Hung-Chieh Fang, Hsuan-Tien Lin, Irwin King et al.

Federated Unsupervised Learning (FUL) aims to learn expressive representations in federated and self-supervised settings. The quality of representations learned in FUL is usually determined by uniformity, a measure of how uniformly representations are distributed in the embedding space. However, existing solutions perform well in achieving intra-client (local) uniformity for local models while failing to achieve inter-client (global) uniformity after aggregation due to non-IID data distributions and the decentralized nature of FUL. To address this issue, we propose Soft Separation and Distillation (SSD), a novel approach that preserves inter-client uniformity by encouraging client representations to spread toward different directions. This design reduces interference during client model aggregation, thereby improving global uniformity while preserving local representation expressiveness. We further enhance this effect by introducing a projector distillation module to address the discrepancy between loss optimization and representation quality. We evaluate SSD in both cross-silo and cross-device federated settings, demonstrating consistent improvements in representation quality and task performance across various training scenarios. Our results highlight the importance of inter-client uniformity in FUL and establish SSD as an effective solution to this challenge. Project page: https://ssd-uniformity.github.io/

Wei-I Lin, Gang Niu, Hsuan-Tien Lin et al.

Complementary-label Learning (CLL) is a form of weakly supervised learning that trains an ordinary classifier using only complementary labels, which are the classes that certain instances do not belong to. While existing CLL studies typically use novel loss functions or training techniques to solve this problem, few studies focus on how complementary labels collectively provide information to train the ordinary classifier. In this paper, we fill the gap by analyzing the implicit sharing of complementary labels on nearby instances during training. Our analysis reveals that the efficiency of implicit label sharing is closely related to the performance of existing CLL models. Based on this analysis, we propose a novel technique that enhances the sharing efficiency via complementary-label augmentation, which explicitly propagates additional complementary labels to each instance. We carefully design the augmentation process to enrich the data with new and accurate complementary labels, which provide CLL models with fresh and valuable information to enhance the sharing efficiency. We then verify our proposed technique by conducting thorough experiments on both synthetic and real-world datasets. Our results confirm that complementary-label augmentation can systematically improve empirical performance over state-of-the-art CLL models.

Si-An Chen, Chun-Liang Li, Hsuan-Tien Lin

Generative Adversarial Networks (GANs) is a powerful family of models that learn an underlying distribution to generate synthetic data. Many existing studies of GANs focus on improving the realness of the generated image data for visual applications, and few of them concern about improving the quality of the generated data for training other classifiers -- a task known as the model compatibility problem. As a consequence, existing GANs often prefer generating `easier' synthetic data that are far from the boundaries of the classifiers, and refrain from generating near-boundary data, which are known to play an important roles in training the classifiers. To improve GAN in terms of model compatibility, we propose Boundary-Calibration GANs (BCGANs), which leverage the boundary information from a set of pre-trained classifiers using the original data. In particular, we introduce an auxiliary Boundary-Calibration loss (BC-loss) into the generator of GAN to match the statistics between the posterior distributions of original data and generated data with respect to the boundaries of the pre-trained classifiers. The BC-loss is provably unbiased and can be easily coupled with different GAN variants to improve their model compatibility. Experimental results demonstrate that BCGANs not only generate realistic images like original GANs but also achieves superior model compatibility than the original GANs.

Cheng-Yu Hsieh, Wei-I Lin, Miao Xu et al.

The goal of multi-label learning (MLL) is to associate a given instance with its relevant labels from a set of concepts. Previous works of MLL mainly focused on the setting where the concept set is assumed to be fixed, while many real-world applications require introducing new concepts into the set to meet new demands. One common need is to refine the original coarse concepts and split them into finer-grained ones, where the refinement process typically begins with limited labeled data for the finer-grained concepts. To address the need, we formalize the problem into a special weakly supervised MLL problem to not only learn the fine-grained concepts efficiently but also allow interactive queries to strategically collect more informative annotations to further improve the classifier. The key idea within our approach is to learn to assign pseudo-labels to the unlabeled entries, and in turn leverage the pseudo-labels to train the underlying classifier and to inform a better query strategy. Experimental results demonstrate that our pseudo-label approach is able to accurately recover the missing ground truth, boosting the prediction performance significantly over the baseline methods and facilitating a competitive active learning strategy.

Ching-Yuan Bai, Hsuan-Tien Lin, Colin Raffel et al.

Many recent developments on generative models for natural images have relied on heuristically-motivated metrics that can be easily gamed by memorizing a small sample from the true distribution or training a model directly to improve the metric. In this work, we critically evaluate the gameability of these metrics by designing and deploying a generative modeling competition. Our competition received over 11000 submitted models. The competitiveness between participants allowed us to investigate both intentional and unintentional memorization in generative modeling. To detect intentional memorization, we propose the ``Memorization-Informed Fréchet Inception Distance'' (MiFID) as a new memorization-aware metric and design benchmark procedures to ensure that winning submissions made genuine improvements in perceptual quality. Furthermore, we manually inspect the code for the 1000 top-performing models to understand and label different forms of memorization. Our analysis reveals that unintentional memorization is a serious and common issue in popular generative models. The generated images and our memorization labels of those models as well as code to compute MiFID are released to facilitate future studies on benchmarking generative models.

Ashesh, Buo-Fu Chen, Treng-Shi Huang et al.

Precipitation nowcasting is an important task for weather forecasting. Many recent works aim to predict the high rainfall events more accurately with the help of deep learning techniques, but such events are relatively rare. The rarity is often addressed by formulations that re-weight the rare events. Somehow such a formulation carries a side effect of making "blurry" predictions in low rainfall regions and cannot convince meteorologists to trust its practical usability. We fix the trust issue by introducing a discriminator that encourages the prediction model to generate realistic rain-maps without sacrificing predictive accuracy. Furthermore, we extend the nowcasting time frame from one hour to three hours to further address the needs from meteorologists. The extension is based on consecutive attentions across different hours. We propose a new deep learning model for precipitation nowcasting that includes both the discrimination and attention techniques. The model is examined on a newly-built benchmark dataset that contains both radar data and actual rain data. The benchmark, which will be publicly released, not only establishes the superiority of the proposed model, but also is expected to encourage future research on precipitation nowcasting.

Michelle Yuan, Hsuan-Tien Lin, Jordan Boyd-Graber

Active learning strives to reduce annotation costs by choosing the most critical examples to label. Typically, the active learning strategy is contingent on the classification model. For instance, uncertainty sampling depends on poorly calibrated model confidence scores. In the cold-start setting, active learning is impractical because of model instability and data scarcity. Fortunately, modern NLP provides an additional source of information: pre-trained language models. The pre-training loss can find examples that surprise the model and should be labeled for efficient fine-tuning. Therefore, we treat the language modeling loss as a proxy for classification uncertainty. With BERT, we develop a simple strategy based on the masked language modeling loss that minimizes labeling costs for text classification. Compared to other baselines, our approach reaches higher accuracy within less sampling iterations and computation time.

Yu-Ting Chou, Gang Niu, Hsuan-Tien Lin et al.

In weakly supervised learning, unbiased risk estimator(URE) is a powerful tool for training classifiers when training and test data are drawn from different distributions. Nevertheless, UREs lead to overfitting in many problem settings when the models are complex like deep networks. In this paper, we investigate reasons for such overfitting by studying a weakly supervised problem called learning with complementary labels. We argue the quality of gradient estimation matters more in risk minimization. Theoretically, we show that a URE gives an unbiased gradient estimator(UGE). Practically, however, UGEs may suffer from huge variance, which causes empirical gradients to be usually far away from true gradients during minimization. To this end, we propose a novel surrogate complementary loss(SCL) framework that trades zero bias with reduced variance and makes empirical gradients more aligned with true gradients in the direction. Thanks to this characteristic, SCL successfully mitigates the overfitting issue and improves URE-based methods.

Chi-Chang Lee, Yu-Chen Lin, Hsuan-Tien Lin et al.

Numerous noise adaptation techniques have been proposed to fine-tune deep-learning models in speech enhancement (SE) for mismatched noise environments. Nevertheless, adaptation to a new environment may lead to catastrophic forgetting of the previously learned environments. The catastrophic forgetting issue degrades the performance of SE in real-world embedded devices, which often revisit previous noise environments. The nature of embedded devices does not allow solving the issue with additional storage of all pre-trained models or earlier training data. In this paper, we propose a regularization-based incremental learning SE (SERIL) strategy, complementing existing noise adaptation strategies without using additional storage. With a regularization constraint, the parameters are updated to the new noise environment while retaining the knowledge of the previous noise environments. The experimental results show that, when faced with a new noise domain, the SERIL model outperforms the unadapted SE model. Meanwhile, compared with the current adaptive technique based on fine-tuning, the SERIL model can reduce the forgetting of previous noise environments by 52%. The results verify that the SERIL model can effectively adjust itself to new noise environments while overcoming the catastrophic forgetting issue. The results make SERIL a favorable choice for real-world SE applications, where the noise environment changes frequently.

Kuen-Han Tsai, Hsuan-Tien Lin

The problem of learning from label proportions (LLP) involves training classifiers with weak labels on bags of instances, rather than strong labels on individual instances. The weak labels only contain the label proportion of each bag. The LLP problem is important for many practical applications that only allow label proportions to be collected because of data privacy or annotation cost, and has recently received lots of research attention. Most existing works focus on extending supervised learning models to solve the LLP problem, but the weak learning nature makes it hard to further improve LLP performance with a supervised angle. In this paper, we take a different angle from semi-supervised learning. In particular, we propose a novel model inspired by consistency regularization, a popular concept in semi-supervised learning that encourages the model to produce a decision boundary that better describes the data manifold. With the introduction of consistency regularization, we further extend our study to non-uniform bag-generation and validation-based parameter-selection procedures that better match practical needs. Experiments not only justify that LLP with consistency regularization achieves superior performance, but also demonstrate the practical usability of the proposed procedures.

Ching-Yuan Bai, Buo-Fu Chen, Hsuan-Tien Lin

Rapid intensification (RI) of tropical cyclones often causes major destruction to human civilization due to short response time. It is an important yet challenging task to accurately predict this kind of extreme weather event in advance. Traditionally, meteorologists tackle the task with human-driven feature extraction and predictor correction procedures. Nevertheless, these procedures do not leverage the power of modern machine learning models and abundant sensor data, such as satellite images. In addition, the human-driven nature of such an approach makes it difficult to reproduce and benchmark prediction models. In this study, we build a benchmark for RI prediction using only satellite images, which are underutilized in traditional techniques. The benchmark follows conventional data science practices, making it easier for data scientists to contribute to RI prediction. We demonstrate the usefulness of the benchmark by designing a domain-inspired spatiotemporal deep learning model. The results showcase the promising performance of deep learning in solving complex meteorological problems such as RI prediction.

Si-An Chen, Voot Tangkaratt, Hsuan-Tien Lin et al.

Recent research has shown that although Reinforcement Learning (RL) can benefit from expert demonstration, it usually takes considerable efforts to obtain enough demonstration. The efforts prevent training decent RL agents with expert demonstration in practice. In this work, we propose Active Reinforcement Learning with Demonstration (ARLD), a new framework to streamline RL in terms of demonstration efforts by allowing the RL agent to query for demonstration actively during training. Under the framework, we propose Active Deep Q-Network, a novel query strategy which adapts to the dynamically-changing distributions during the RL training process by estimating the uncertainty of recent states. The expert demonstration data within Active DQN are then utilized by optimizing supervised max-margin loss in addition to temporal difference loss within usual DQN training. We propose two methods of estimating the uncertainty based on two state-of-the-art DQN models, namely the divergence of bootstrapped DQN and the variance of noisy DQN. The empirical results validate that both methods not only learn faster than other passive expert demonstration methods with the same amount of demonstration and but also reach super-expert level of performance across four different tasks.

Yao-Yuan Yang, Yi-An Lin, Hong-Min Chu et al.

Multi-label classification (MLC) is an important class of machine learning problems that come with a wide spectrum of applications, each demanding a possibly different evaluation criterion. When solving the MLC problems, we generally expect the learning algorithm to take the hidden correlation of the labels into account to improve the prediction performance. Extracting the hidden correlation is generally a challenging task. In this work, we propose a novel deep learning framework to better extract the hidden correlation with the help of the memory structure within recurrent neural networks. The memory stores the temporary guesses on the labels and effectively allows the framework to rethink about the goodness and correlation of the guesses before making the final prediction. Furthermore, the rethinking process makes it easy to adapt to different evaluation criteria to match real-world application needs. In particular, the framework can be trained in an end-to-end style with respect to any given MLC evaluation criteria. The end-to-end design can be seamlessly combined with other deep learning techniques to conquer challenging MLC problems like image tagging. Experimental results across many real-world data sets justify that the rethinking framework indeed improves MLC performance across different evaluation criteria and leads to superior performance over state-of-the-art MLC algorithms.

Yong-Siang Shih, Kai-Yueh Chang, Hsuan-Tien Lin et al.

Compatibility between items, such as clothes and shoes, is a major factor among customer's purchasing decisions. However, learning "compatibility" is challenging due to (1) broader notions of compatibility than those of similarity, (2) the asymmetric nature of compatibility, and (3) only a small set of compatible and incompatible items are observed. We propose an end-to-end trainable system to embed each item into a latent vector and project a query item into K compatible prototypes in the same space. These prototypes reflect the broad notions of compatibility. We refer to both the embedding and prototypes as "Compatibility Family". In our learned space, we introduce a novel Projected Compatibility Distance (PCD) function which is differentiable and ensures diversity by aiming for at least one prototype to be close to a compatible item, whereas none of the prototypes are close to an incompatible item. We evaluate our system on a toy dataset, two Amazon product datasets, and Polyvore outfit dataset. Our method consistently achieves state-of-the-art performance. Finally, we show that we can visualize the candidate compatible prototypes using a Metric-regularized Conditional Generative Adversarial Network (MrCGAN), where the input is a projected prototype and the output is a generated image of a compatible item. We ask human evaluators to judge the relative compatibility between our generated images and images generated by CGANs conditioned directly on query items. Our generated images are significantly preferred, with roughly twice the number of votes as others.

Hong-Min Chu, Kuan-Hao Huang, Hsuan-Tien Lin

We study multi-label classification (MLC) with three important real-world issues: online updating, label space dimensional reduction (LSDR), and cost-sensitivity. Current MLC algorithms have not been designed to address these three issues simultaneously. In this paper, we propose a novel algorithm, cost-sensitive dynamic principal projection (CS-DPP) that resolves all three issues. The foundation of CS-DPP is an online LSDR framework derived from a leading LSDR algorithm. In particular, CS-DPP is equipped with an efficient online dimension reducer motivated by matrix stochastic gradient, and establishes its theoretical backbone when coupled with a carefully-designed online regression learner. In addition, CS-DPP embeds the cost information into label weights to achieve cost-sensitivity along with theoretical guarantees. Experimental results verify that CS-DPP achieves better practical performance than current MLC algorithms across different evaluation criteria, and demonstrate the importance of resolving the three issues simultaneously.

Te-Kang Jan, Da-Wei Wang, Chi-Hung Lin et al.

Many real-world data mining applications need varying cost for different types of classification errors and thus call for cost-sensitive classification algorithms. Existing algorithms for cost-sensitive classification are successful in terms of minimizing the cost, but can result in a high error rate as the trade-off. The high error rate holds back the practical use of those algorithms. In this paper, we propose a novel cost-sensitive classification methodology that takes both the cost and the error rate into account. The methodology, called soft cost-sensitive classification, is established from a multicriteria optimization problem of the cost and the error rate, and can be viewed as regularizing cost-sensitive classification with the error rate. The simple methodology allows immediate improvements of existing cost-sensitive classification algorithms. Experiments on the benchmark and the real-world data sets show that our proposed methodology indeed achieves lower test error rates and similar (sometimes lower) test costs than existing cost-sensitive classification algorithms. We also demonstrate that the methodology can be extended for considering the weighted error rate instead of the original error rate. This extension is useful for tackling unbalanced classification problems.

Wei-Yuan Shen, Hsuan-Tien Lin

Bipartite ranking is a fundamental ranking problem that learns to order relevant instances ahead of irrelevant ones. The pair-wise approach for bi-partite ranking construct a quadratic number of pairs to solve the problem, which is infeasible for large-scale data sets. The point-wise approach, albeit more efficient, often results in inferior performance. That is, it is difficult to conduct bipartite ranking accurately and efficiently at the same time. In this paper, we develop a novel active sampling scheme within the pair-wise approach to conduct bipartite ranking efficiently. The scheme is inspired from active learning and can reach a competitive ranking performance while focusing only on a small subset of the many pairs during training. Moreover, we propose a general Combined Ranking and Classification (CRC) framework to accurately conduct bipartite ranking. The framework unifies point-wise and pair-wise approaches and is simply based on the idea of treating each instance point as a pseudo-pair. Experiments on 14 real-word large-scale data sets demonstrate that the proposed algorithm of Active Sampling within CRC, when coupled with a linear Support Vector Machine, usually outperforms state-of-the-art point-wise and pair-wise ranking approaches in terms of both accuracy and efficiency.

Yao-Yuan Yang, Kuan-Hao Huang, Chih-Wei Chang et al.

Label space expansion for multi-label classification (MLC) is a methodology that encodes the original label vectors to higher dimensional codes before training and decodes the predicted codes back to the label vectors during testing. The methodology has been demonstrated to improve the performance of MLC algorithms when coupled with off-the-shelf error-correcting codes for encoding and decoding. Nevertheless, such a coding scheme can be complicated to implement, and cannot easily satisfy a common application need of cost-sensitive MLC---adapting to different evaluation criteria of interest. In this work, we show that a simpler coding scheme based on the concept of a reference pair of label vectors achieves cost-sensitivity more naturally. In particular, our proposed cost-sensitive reference pair encoding (CSRPE) algorithm contains cluster-based encoding, weight-based training and voting-based decoding steps, all utilizing the cost information. Furthermore, we leverage the cost information embedded in the code space of CSRPE to propose a novel active learning algorithm for cost-sensitive MLC. Extensive experimental results verify that CSRPE performs better than state-of-the-art algorithms across different MLC criteria. The results also demonstrate that the CSRPE-backed active learning algorithm is superior to existing algorithms for active MLC, and further justify the usefulness of CSRPE.

Yu-An Chung, Shao-Wen Yang, Hsuan-Tien Lin

While deep neural networks have succeeded in several visual applications, such as object recognition, detection, and localization, by reaching very high classification accuracies, it is important to note that many real-world applications demand varying costs for different types of misclassification errors, thus requiring cost-sensitive classification algorithms. Current models of deep neural networks for cost-sensitive classification are restricted to some specific network structures and limited depth. In this paper, we propose a novel framework that can be applied to deep neural networks with any structure to facilitate their learning of meaningful representations for cost-sensitive classification problems. Furthermore, the framework allows end-to-end training of deeper networks directly. The framework is designed by augmenting auxiliary neurons to the output of each hidden layer for layer-wise cost estimation, and including the total estimation loss within the optimization objective. Experimental results on public benchmark visual data sets with two cost information settings demonstrate that the proposed framework outperforms state-of-the-art cost-sensitive deep learning models.

Hong-Min Chu, Hsuan-Tien Lin

Active learning is an important machine learning problem in reducing the human labeling effort. Current active learning strategies are designed from human knowledge, and are applied on each dataset in an immutable manner. In other words, experience about the usefulness of strategies cannot be updated and transferred to improve active learning on other datasets. This paper initiates a pioneering study on whether active learning experience can be transferred. We first propose a novel active learning model that linearly aggregates existing strategies. The linear weights can then be used to represent the active learning experience. We equip the model with the popular linear upper- confidence-bound (LinUCB) algorithm for contextual bandit to update the weights. Finally, we extend our model to transfer the experience across datasets with the technique of biased regularization. Empirical studies demonstrate that the learned experience not only is competitive with existing strategies on most single datasets, but also can be transferred across datasets to improve the performance on future learning tasks.

Chih-Kuan Yeh, Hsuan-Tien Lin

Bridge is among the zero-sum games for which artificial intelligence has not yet outperformed expert human players. The main difficulty lies in the bidding phase of bridge, which requires cooperative decision making under partial information. Existing artificial intelligence systems for bridge bidding rely on and are thus restricted by human-designed bidding systems or features. In this work, we propose a pioneering bridge bidding system without the aid of human domain knowledge. The system is based on a novel deep reinforcement learning model, which extracts sophisticated features and learns to bid automatically based on raw card data. The model includes an upper-confidence-bound algorithm and additional techniques to achieve a balance between exploration and exploitation. Our experiments validate the promising performance of our proposed model. In particular, the model advances from having no knowledge about bidding to achieving superior performance when compared with a champion-winning computer bridge program that implements a human-designed bidding system.

Kuan-Hao Huang, Hsuan-Tien Lin

Label embedding (LE) is an important family of multi-label classification algorithms that digest the label information jointly for better performance. Different real-world applications evaluate performance by different cost functions of interest. Current LE algorithms often aim to optimize one specific cost function, but they can suffer from bad performance with respect to other cost functions. In this paper, we resolve the performance issue by proposing a novel cost-sensitive LE algorithm that takes the cost function of interest into account. The proposed algorithm, cost-sensitive label embedding with multidimensional scaling (CLEMS), approximates the cost information with the distances of the embedded vectors by using the classic multidimensional scaling approach for manifold learning. CLEMS is able to deal with both symmetric and asymmetric cost functions, and effectively makes cost-sensitive decisions by nearest-neighbor decoding within the embedded vectors. We derive theoretical results that justify how CLEMS achieves the desired cost-sensitivity. Furthermore, extensive experimental results demonstrate that CLEMS is significantly better than a wide spectrum of existing LE algorithms and state-of-the-art cost-sensitive algorithms across different cost functions.

Yu-An Chung, Hsuan-Tien Lin, Shao-Wen Yang

Deep learning has been one of the most prominent machine learning techniques nowadays, being the state-of-the-art on a broad range of applications where automatic feature extraction is needed. Many such applications also demand varying costs for different types of mis-classification errors, but it is not clear whether or how such cost information can be incorporated into deep learning to improve performance. In this work, we propose a novel cost-aware algorithm that takes into account the cost information into not only the training stage but also the pre-training stage of deep learning. The approach allows deep learning to conduct automatic feature extraction with the cost information effectively. Extensive experimental results demonstrate that the proposed approach outperforms other deep learning models that do not digest the cost information in the pre-training stage.

Chun-Liang Li, Hsuan-Tien Lin, Chi-Jen Lu

We study the problem of recovering the subspace spanned by the first $k$ principal components of $d$-dimensional data under the streaming setting, with a memory bound of $O(kd)$. Two families of algorithms are known for this problem. The first family is based on the framework of stochastic gradient descent. Nevertheless, the convergence rate of the family can be seriously affected by the learning rate of the descent steps and deserves more serious study. The second family is based on the power method over blocks of data, but setting the block size for its existing algorithms is not an easy task. In this paper, we analyze the convergence rate of a representative algorithm with decayed learning rate (Oja and Karhunen, 1985) in the first family for the general $k>1$ case. Moreover, we propose a novel algorithm for the second family that sets the block sizes automatically and dynamically with faster convergence rate. We then conduct empirical studies that fairly compare the two families on real-world data. The studies reveal the advantages and disadvantages of these two families.

Shang-Tse Chen, Hsuan-Tien Lin, Chi-Jen Lu

We study the task of online boosting--combining online weak learners into an online strong learner. While batch boosting has a sound theoretical foundation, online boosting deserves more study from the theoretical perspective. In this paper, we carefully compare the differences between online and batch boosting, and propose a novel and reasonable assumption for the online weak learner. Based on the assumption, we design an online boosting algorithm with a strong theoretical guarantee by adapting from the offline SmoothBoost algorithm that matches the assumption closely. We further tackle the task of deciding the number of weak learners using established theoretical results for online convex programming and predicting with expert advice. Experiments on real-world data sets demonstrate that the proposed algorithm compares favorably with existing online boosting algorithms.