查看更多>>摘要：Conversational recommender systems (CRSs) provide personalised recommendations by strategically querying attributes matching users’ preferences. However, this process suffers from confounding effects of time and user attributes, as users’ preferences naturally evolve over time and differ among similar users due to their unique attributes. These confounding effects distort user behaviors’ causal drivers, challenging CRSs in learning users’ true preferences and generalizable patterns. Recently, causal inference provides principled tools to clarify cause-effect relations in data, offering a promising way to address such confounding effects. In this context, we introduce Causal Conversational Recommender (CCR), which applies causal inference to model the causality between user behaviors and time/user attribute, enabling deeper understanding of user behaviors’ causal drivers. First, CCR employs stratification and matching to ensure attribute asked per round is independent from time and user attributes, mitigating their confounding effects. Following that, we apply the Average Treatment Effect (ATE) to quantify the unbiased causal impact of each unasked attribute on user preferences, identifying the attribute with the highest ATE per round as the causal-based attribute, i.e., causal driver of user behaviour. Finally, CCR iteratively refines user preferences through feedback on causal-based attributes. Extensive experiments verified CCR's robustness and personalization.
查看更多>>摘要：The class imbalance problem can cause classifiers to be biased toward the majority class and inclined to generate incorrect predictions. While existing studies have proposed numerous oversampling methods to alleviate class imbalance by generating extra minority class samples, these methods still have some inherent weaknesses and make the generated samples less informative. This study proposes a novel over-sampling method named the Expandable Borderline Smote (EB-Smote), which can address the weaknesses of existing over-sampling methods and generate more informative synthetic samples. In EB-Smote, not only minority class but also majority class is oversampled, and the synthetic samples are generated in the area between the selected minority and majority samples, which are close to the borderlines of their respective classes. EB-Smote can generate more informative samples by expanding the borderlines of minority and majority classes toward the actual decision boundary. Based on 27 imbalanced datasets and commonly used machine learning models, the experimental results demonstrate that EB-Smote significantly outperforms the other 8 existing oversampling methods. This study can provide theoretical guidance and practical recommendations to solve the crucial class imbalance problem in classification tasks.
查看更多>>摘要：Bandit online multiclass prediction plays an important role in many real-world applications. In this paper, we propose a unified Bandit Online Multiclass Prediction (BOMP) framework. This framework is based on our proposed margin-based gradient descent approach. Its update step provides an unbiased estimate of the surrogate loss gradient and has a lower variance than existing methods. It also enables our algorithms to update even for incorrect predictions by penalizing the wrong classes. The link function of the framework can evolve over time, gradually incorporating online data information including second-order information into the potential functions. Based on the proposed framework, we investigate first-order and second-order bandit online multiclass prediction algorithms. Theoretical analysis demonstrates the superiority of our proposed update rule and bandit online multiclass prediction framework. Finally, we compare our proposed first-order and second-order bandit online multiclass prediction algorithms with several state-of-the-art methods on two synthetic and four real-world datasets. The encouraging results show that our proposed algorithms significantly outperform state-of-the-art techniques.
查看更多>>摘要：Urban spatio-temporal prediction is crucial for informed decision-making, such as traffic management, resource optimization, and emergency response. Despite remarkable breakthroughs in pretrained natural language models that enable one model to handle diverse tasks, a universal solution for spatio-temporal prediction remains challenging. Existing prediction approaches are typically tailored for specific spatio-temporal scenarios, requiring task-specific model designs and extensive domain-specific training data. In this study, we introduce UniST, a universal model designed for general urban spatio-temporal prediction across a wide range of scenarios. Inspired by large language models, UniST achieves success through: (i) utilizing diverse spatio-temporal data from different scenarios, (ii) effective pre-training to capture complex spatio-temporal dynamics, (iii) knowledge-guided prompts to enhance generalization capabilities. These designs together unlock the potential of building a universal model for various scenarios. Extensive experiments on more than 20 spatio-temporal scenarios, including grid-based data and graph-based data, demonstrate UniST’s efficacy in advancing state-of-the-art performance, especially in few-shot and zero-shot prediction.
查看更多>>摘要：Graph Convolutional Networks (GCNs) have demonstrated remarkable success in various graph-related tasks. However, recent studies show that GCNs are vulnerable to adversarial attacks on graph structures. Therefore, how to defend against such attacks has become a popular research topic. The current common defense methods face two main limitations: (1) From the data perspective, it may lead to suboptimal results since the structural information is ignored when distinguishing the perturbed edges. (2) From the model perspective, the defenders rely on the low-pass filter of the GCN, which is vulnerable during message passing. To overcome these limitations, this paper analyzes the characteristics of perturbed edges, and based on this we propose a robust defense framework, REDE, to generate the adaptive Reliable Defense graph for multi-channel robust GCN. REDE first uses feature similarity and structure difference to discriminate perturbed edges and generates the defense graph by pruning them. Then REDE designs a multi-channel GCN, which can separately capture the information of different edges and high-order neighbors utilizing different frequency components. Leveraging this capability, the defense graph is adaptively updated at each layer, enhancing its reliability and improving prediction accuracy. Extensive experiments on four benchmark datasets demonstrate the enhanced performance and robustness of our proposed REDE over the state-of-the-art defense methods.
查看更多>>摘要：Motif discovery is a critical operation for analyzing series data in many applications. Recent works demonstrate the importance of finding motifs with Dynamic Time Warping. However, existing algorithms spend most of their time in computing lower bounds of Dynamic Time Warping to filter out the unpromising candidates. Specifically, the time complexity for computing these lower bounds is $O(L)$ for each pair of subsequences, where $L$ is the length of the motif (subsequences). This paper proposes two new lower bounds, called $LB_{f}$ and $LB_{M}$, both of them only cost amortized $O(1)$ time for each pair of subsequences. On real datasets, the proposed lower bounds are at least one magnitude faster than the state-of-the-art lower bounds used in motif discovery while still keeping satisfying effectiveness. Based on these faster lower bounds, this paper designs an efficient motif discovery algorithm that significantly reduces the cost of lower bounds. The experiments conducted on real datasets show the proposed algorithm is 5.6 times faster than the state-of-the-art algorithms on average.
Mouxiang ChenHan FuChenghao LiuXiaoyun Joy Wang...
2253-2265页
查看更多>>摘要：Prompt tuning for pre-trained language models (PLMs) has been an effective approach for few-shot text classification. To make a prediction, a typical prompt tuning method employs a template wrapping the input text into a cloze question, and a verbalizer mapping the output embedding to labels. However, current methods typically depend on handcrafted templates and verbalizers, which require much domain-specific prior knowledge by human efforts. In this work, we investigate how to build a good human-free prompt tuning using soft prompt templates and soft verbalizers, which can be learned directly from data. To address the challenge of data scarcity, we integrate a set of trainable bases for sentence representation to transfer the contextual information into a low-dimensional space. By jointly pre-training the soft prompts and the bases using contrastive learning, the projection space can catch critical semantics at the sentence level, which could be transferred to various downstream tasks. To better bridge the gap between downstream tasks and the pre-training procedure, we formulate the few-shot classification tasks as another contrastive learning problem. We name this Jointly Pretrained Template and Verbalizer (JPTV). Extensive experiments show that this human-free prompt tuning can achieve comparable or even better performance than manual prompt tuning.
查看更多>>摘要：Federated learning (FL), a decentralized machine learning approach, offers great performance while alleviating autonomy and confidentiality concerns. Despite FL’s popularity, how to deal with missing values in a federated manner is not well understood. In this work, we initiate a study of federated imputation of missing values, particularly in complex scenarios, where missing data heterogeneity exists and the state-of-the-art (SOTA) approaches for federated imputation suffer from significant loss in imputation quality. We propose Cafe, a personalized FL approach for missing data imputation. Cafe is inspired from the observation that heterogeneity can induce differences in observable and missing data distribution across clients, and that these differences can be leveraged to improve the imputation quality. Cafe computes personalized weights that are automatically calibrated for the level of heterogeneity, which can remain unknown, to develop personalized imputation models for each client. An extensive empirical evaluation over a variety of settings demonstrates that Cafe matches the performance of SOTA baselines in homogeneous settings while significantly outperforming the baselines in heterogeneous settings.
查看更多>>摘要：Federated Class Incremental Learning (FCIL) has emerged as a new paradigm due to its applicability in real-world scenarios. In FCIL, clients continuously generate new data with unseen class labels and do not share local data due to privacy restrictions, and each client’s class distribution evolves dynamically and independently. However, existing work still faces two significant challenges. Firstly, current methods lack a better balance between maintaining sound anti-forgetting effects over old data (stability) and ensuring good adaptability for new tasks (plasticity). Secondly, some FCIL methods overlook that the incremental data will also have a non-identical label distribution, leading to poor performance. This paper proposes CGoFed, which includes relax-constrained gradient update and cross-task gradient regularization modules. The relax-constrained gradient update prevents forgetting the knowledge about old data while quickly adapting to the new data by constraining the gradient update direction to a gradient space that minimizes interference with historical tasks. The cross-task gradient regularization also finds applicable historical models from other clients and trains a personalized global model to address the non-identical label distribution problem. The results demonstrate that the CGoFed performs well in alleviating catastrophic forgetting and improves model performance by 8% -23% compared with the SOTA comparison method.
查看更多>>摘要：Open information extraction (OIE) methods extract plenty of OIE triples $ $noun phrase, relation phrase, noun phrase$ $ from unstructured text, which compose large open knowledge bases (OKBs). Noun phrases and relation phrases in such OKBs are not canonicalized, which leads to scattered and redundant facts. It is found that two views of knowledge (i.e., a fact view based on the fact triple and a context view based on the fact triple's source context) provide complementary information that is vital to the task of OKB canonicalization, which clusters synonymous noun phrases and relation phrases into the same group and assigns them unique identifiers. In order to leverage these two views of knowledge jointly, we propose CMVC+, a novel unsupervised framework for canonicalizing OKBs without the need for manually annotated labels. Specifically, we propose a multi-view CHF K-Means clustering algorithm to mutually reinforce the clustering of view-specific embeddings learned from each view by considering the clustering quality in a fine-grained manner. Furthermore, we propose a novel contrastive learning module to refine the learned view-specific embeddings and further enhance the canonicalization performance. We demonstrate the superiority of our framework through extensive experiments on multiple real-world OKB data sets against state-of-the-art methods.
NETL
NSTL
IEEE