IJCAI 2024 Papers

Although Hierarchical Federated Learning (HFL) utilizes edge servers (ESs) to alleviate communication burdens, its model performance will be degraded by non-IID data and limited communication resources. Current works often assume that data is uniformly distributed, which however contradicts the heterogeneity of IoT. Solutions involving additional model training to check the data distribution inevitably increase computational costs and the risk of privacy leakage. The challenges in solving these issues are how to reduce the impact of non-IID data without involving raw data, and how to rationalize the communication resource allocation for addressing straggler problem. To tackle these challenges, we propose a novel optimization method based on coaLition formation gamE and grAdient Projection, called LEAP. Specifically, we combine edge data distribution with coalition formation game innovatively to adjust the correlations between clients and ESs dynamically, ensuring optimal correlations. We further capture the client heterogeneity to achieve the rational bandwidth allocation from coalition perception and determine the optimal transmission power within specified delay constraints at the client level. Experimental results on four real datasets show that LEAP is able to achieve 20.62% improvement in model accuracy compared to the state-of-the-art baselines. Moreover, LEAP effectively reduces transmission energy consumption by at least about 2.24 times.

This paper describes the KnowledgeHub tool, a scientific literature Information Extraction (IE) and Question Answering (QA) pipeline. This is achieved by supporting the ingestion of PDF documents that are converted to text and structured representations. An ontology can then be constructed where a user defines the types of entities and relationships they want to capture. A browser-based annotation tool enables annotating the contents of the PDF documents according to the ontology. Named Entity Recognition (NER) and Relation Classification (RC) models can be trained on the resulting annotations and can be used to annotate the unannotated portion of the documents. A knowledge graph is constructed from these entity and relation triples which can be queried to obtain insights from the data. Furthermore, we integrate a suite of Large Language Models (LLMs) that can be used for QA and summarisation that is grounded in the included documents. KnowledgeHub is a unique tool that supports annotation, IE and QA, which gives the user full insight into the knowledge discovery pipeline.

Federated Learning (FL) enables the training of Deep Learning models without centrally collecting possibly sensitive raw data. The most used algorithms for FL are parameter-averaging based schemes (e.g., Federated Averaging) that, however, have well known limits, i.e., model homogeneity, high communication cost, poor performance in presence of heterogeneous data distributions. Federated adaptations of regular Knowledge Distillation (KD) can solve or mitigate the weaknesses of parameter-averaging FL algorithms while possibly introducing other trade-offs. In this article, we originally present a focused review of the state-of-the-art KD-based algorithms specifically tailored for FL, by providing both a novel classification of the existing approaches and a detailed technical description of their pros, cons, and tradeoffs.

Knowledge compilation has proven effective in (weighted) model counting, uniquely supporting incrementality and checkability. For incrementality, compiling an input formula once suffices to answer multiple queries, thus reducing the total solving effort. For checkability, the compiled formula is amenable to producing machine-checkable proofs for verification, thus strengthening the solver’s reliability. In this work, we extend knowledge compilation from model counting to stochastic Boolean satisfiability (SSAT) solving by generalizing the dec-DNNF representation to accommodate the SSAT quantifier structure and integrate it into SharpSSAT, a state-of-the-art SSAT solver. We further study proof generation from the compiled representation and extend CPOG, a certified model-counting toolchain, to generate proofs for certifying the results of SharpSSAT. Experimental results show the benefits of the proposed knowledge compilation approach for SSAT in sharing computation efforts for multiple queries and producing checkable dec-DNNF logs with negligible overhead.

Graph neural networks (GNNs) for graph classification or representation learning require a pooling operation to convert the nodes' embeddings of each graph to a vector as the graph-level representation and the operation has a significant impact on model accuracy. The paper presents a novel graph pooling method called Kernel Readout (KerRead). KerRead maps the node embeddings from the sample space with limited nodes to an augmented sample space with infinite nodes, and then calculates the inner product between some learnable adaptive centers and the augmented node embeddings, which forms a final graph-level feature vector. We apply the proposed strategy to six supervised and two unsupervised graph neural networks such as GCN, GIN, GUNet, InfoGraph, and GraphCL, and the experiments on eight benchmark datasets show that the proposed readout outperforms classical pooling methods such as Sum and seven state-of-the-art pooling methods such as SRead and Janossy GRU. Code and Appendix are both available at https://github.com/jiajunCAU/KerRead.

Open-World Object Detection (OWOD) has garnered widespread attention due to its ability to recall unannotated objects. Existing works generate pseudo-labels for the model using heuristic priors, which limits the model’s performance. In this paper, we leverage the knowledge of the large-scale visual model to provide supervision for unknown categories. Specifically, we use the Segment Anything Model (SAM) to generate raw pseudo-labels for potential objects and refine them through Intersection over Union (IOU) and the shortest bounding box side length. Nevertheless, the abundance of pseudo-labels still exacerbates the competition issue in the one-to-many label assignment. To address this, we propose the Dual Matching Label Assignment (DMLA) strategy. Furthermore, we propose the Class-Awareness Neutralizer (CAN) to reduce the model’s bias towards known categories. Evaluation results on open-world object detection benchmarks, including MS COCO and Pascal VOC, show that our method achieves nearly 200% the unknown recall rate of previous state-of-the-art (SOTA) methods, reaching 41.5 U-Recall. Additionally, our approach does not add any extra parameters, maintaining the inference speed advantage of Faster R-CNN, leading the SOTA methods based on deformable DETR at a speed of over 10 FPS. Our code is available at https://github.com/xxyzll/KTCN.

Large language models (LLMs) encounter challenges such as hallucination and factual errors in knowledge-intensive tasks. One the one hand, LLMs sometimes struggle to generate reliable answers based on the black-box parametric knowledge, due to the lack of responsible knowledge. Moreover, fragmented knowledge facts extracted by knowledge retrievers fail to provide explicit and coherent reasoning paths for improving LLM reasoning. To address these challenges, we propose KG-CoT, a novel knowledge-augmented paradigm that leverages a small-scale step-by-step graph reasoning model to reason over knowledge graphs (KGs) and utilizes a reasoning path generation method to generate chains of reasoning with high confidence for large-scale LLMs. Extensive experiments demonstrate that our KG-CoT significantly improves the performance of LLMs on knowledge-intensive question answering tasks, such as multi-hop, single-hop, and open-domain question answering benchmarks, without fine-tuning LLMs. KG-CoT outperforms the CoT prompting as well as prior retrieval-augmented and knowledge base question answering baselines. Moreover, KG-CoT can reduce the number of API calls and cost and generalize to various LLM backbones in a lightweight plug-and-play manner.

Pre-travel recommendation is developed to provide a variety of out-of-town Point-of-Interests (POIs) for users planning to travel away from their hometowns but have not yet decided on their destination. Existing out-of-town recommender systems work on constructing users' latent preferences and inferring travel intentions from their check-in sequences. However, there are still two challenges that hamper the performance of these approaches: i) Users' interactive data (including hometown and out-of-town check-ins) tend to be rare, and while candidate POIs that come from different regions contain various semantic information; ii) The causes for user check-in include not only interest but also conformity, which are easily entangled and overlooked. To fill these gaps, we propose a Knowledge-Driven Disentangled Causal metric learning framework (KDDC) that mitigates interaction data sparsity by enhancing POI semantic representation and considers the distributions of two causes (i.e., conformity and interest) for pre-travel recommendation. Specifically, we pretrain a constructed POI attribute knowledge graph through a segmented interaction method and POI semantic information is aggregated via relational heterogeneity. In addition, we devise a disentangled causal metric learning to model and infer userrelated representations. Extensive experiments on two real-world nationwide datasets display the consistent superiority of our KDDC over state-of-theart baselines.

Exploring the narratives conveyed by fine-art paintings is a challenge in image captioning, where the goal is to generate descriptions that not only precisely represent the visual content but also offer a in-depth interpretation of the artwork's meaning. The task is particularly complex for artwork images due to their diverse interpretations and varied aesthetic principles across different artistic schools and styles. In response to this, we present KALE (Knowledge-Augmented vision-Language model for artwork Elaborations), a novel approach that enhances existing vision-language models by integrating artwork metadata as additional knowledge. KALE incorporates the metadata in two ways: firstly as direct textual input, and secondly through a multimodal heterogeneous knowledge graph. To optimize the learning of graph representations, we introduce a new cross-modal alignment loss that maximizes the similarity between the image and its corresponding metadata. Experimental results demonstrate that KALE achieves strong performance (when evaluated with CIDEr, in particular) over existing state-of-the-art work across several artwork datasets. Source code of the project is available at https://github.com/Yanbei-Jiang/Artwork-Interpretation.

In formal argumentation one aims for intuitive and concise justifications for the acceptance of arguments. Discussion games and dispute trees are established methods to obtain such a justification. However, so far these techniques are based on instantiating the knowledge base into graph-based Dung style abstract argumentation frameworks (AFs). These instantiations are known to produce frameworks with a large number of arguments and thus also yield long discussion games and large dispute trees. To obtain more concise justifications for argument acceptance, we propose to instantiate the knowledge base as an argumentation framework with collective attacks (SETAF). Remarkably, this approach yields smaller frameworks compared to traditional AF instantiation, while exhibiting increased expressive power. We then introduce discussion games and dispute trees tailored to SETAFs, show that they correspond to credulous acceptance w.r.t. the well-known preferred semantics, analyze and tune them w.r.t. the size, and compare the two notions. Finally, we illustrate how our findings apply to assumption-based argumentation.

Different social media are widely used in our daily lives. Inspired by the fact that similar topics have similar propagation characteristics, we mine the implicit knowledge of cascades with similar topics from different platforms to enhance the localization performance for scenarios where limited propagation data leads to the weak learning ability of existing localization models. In this work, we first construct a multiple platform propagation cascade dataset, aligning similar topics from both Twitter and Weibo, and enriching it with user profiles. Leveraging this dataset, we propose a Dual-channel Source Localization Framework (DSLF) for the joint cascades with similar topics. Specifically, a self-loop attention based graph convolutional network is designed to adaptively adjust the neighborhood aggregation scheme of different users with heterogeneous features in the message-passing process. Additionally, a dual-structure based Kullback-Leibler (KL) regularization module is proposed to constrain the latent distribution space of the source probabilities of similar characteristic-level users for a similar topic, enhancing the robustness of the model. Extensive experiments across Twitter and Weibo platforms demonstrate the superiority of the proposed DSLF over the SOTA methods. The code is available at https://github.com/cgao-comp/DSLF.

As an important task in sentiment analysis, joint multimodal aspect sentiment analysis (JMASA) has received increasing attention in recent years. However, previous approaches either i) directly fuse multimodal data without fully exploiting the correlation between multimodal input data, or ii) equally utilize the dependencies of words in the text for sentiment analysis, ignoring the differences in the importance of different words. To address these limitations, we propose a joint multimodal sentiment analysis method based on Aspect Enhancement and Syntactic Adaptive Learning (AESAL). Specifically, we construct an aspect enhancement pre-training task to enable the model to fully learn the correlation of aspects between multimodal input data. In order to capture the differences in the importance of different words in the text, we design a syntactic adaptive learning mechanism. First, we construct different syntactic dependency graphs based on the distance between words to learn global and local information in the text. Second, we use a multi-channel adaptive graph convolutional network to maintain the uniqueness of each modality while fusing the correlations between different modalities. Experimental results on benchmark datasets show that our method outperforms state-of-the-art methods.

Incremental learning is nontrivial due to severe catastrophic forgetting. Although storing a small amount of data on old tasks during incremental learning is a feasible solution, current strategies still do not 1) adequately address the class bias problem, and 2) alleviate the mutual interference between new and old tasks, and 3) consider the problem of class bias within tasks. In light of the above issues, we analyze the cause of class bias in incremental learning, as well as the drawbacks of existing approaches, and propose a joint input and output coordination (JIOC) mechanism to address these issues. This mechanism assigns different weights to different categories of data according to the gradient of the output score, and uses knowledge distillation (KD) to reduce the mutual interference between the outputs of old and new tasks. The proposed mechanism is general and flexible, and can be incorporated into different incremental learning approaches that use memory storage. Extensive experiments show that our mechanism can significantly improve their performance.

In the realm of cross-network tasks, graph domain adaptation is an effective tool due to its ability to transfer abundant labels from nodes in the source domain to those in the target domain. Existing adversarial domain adaptation methods mainly focus on domain-wise alignment. These approaches, while effective in mitigating the marginal distribution shift between the two domains, often ignore the integral aspect of structural alignment, potentially leading to negative transfer. To address this issue, we propose a joint adversarial domain adaptive graph convolutional network (JDA-GCN) that is uniquely augmented with structural graph alignment, so as to enhance the efficacy of knowledge transfer. Specifically, we construct a structural graph to delineate the interconnections among nodes within identical categories across the source and target domains. To further refine node representation, we integrate the local consistency matrix with the global consistency matrix, thereby leveraging the learning of the sub-structure similarity of nodes to enable more robust and effective representation of nodes. Empirical evaluation on diverse real-world datasets substantiates the superiority of our proposed method, marking a significant advancement over existing state-of-the-art graph domain adaptation algorithms.

Large language models (LLMs) have achieved remarkable proficiency on solving diverse problems. However, their generalization ability is not always satisfying and the generalization problem is common for generative transformer models in general. Researchers take basic mathematical tasks like n-digit addition or multiplication as important perspectives for investigating their generalization behaviors. It is observed that when training models on n-digit operations (e.g., additions) in which both input operands are n-digit in length, models generalize successfully on unseen n-digit inputs (in-distribution (ID) generalization), but fail miserably on longer, unseen cases (out-of-distribution (OOD) generalization). We bring this unexplained performance drop into attention and ask whether there is systematic OOD generalization. Towards understanding LLMs, we train various smaller language models which may share the same underlying mechanism. We discover that the strong ID generalization stems from structured representations, while behind the unsatisfying OOD performance, the models still exhibit clear learned algebraic structures. Specifically, these models map unseen OOD inputs to outputs with learned equivalence relations in the ID domain, which we call the equivalence generalization. These findings deepen our knowledge regarding the generalizability of generative models including LLMs, and provide insights into potential avenues for improvement.

Invertible Rescaling Networks (IRNs) and their variants have witnessed remarkable achievements in various image processing tasks like image rescaling. However, we observe that IRNs with deeper networks are difficult to train, thus hindering the representational ability of IRNs. To address this issue, we propose Invertible Residual Rescaling Models (IRRM) for image rescaling by learning a bijection between a high-resolution image and its low-resolution counterpart with a specific distribution. Specifically, we propose IRRM to build a deep network, which contains several Residual Downscaling Modules (RDMs) with long skip connections. Each RDM consists of several Invertible Residual Blocks (IRBs) with short connections. In this way, RDM allows rich low-frequency information to be bypassed by skip connections and forces models to focus on extracting high-frequency information from the image. Extensive experiments show that our IRRM performs significantly better than other state-of-the-art methods with much fewer parameters and complexity. Particularly, our IRRM has respectively PSNR gains of at least 0.3 dB over HCFlow and IRN in the x4 rescaling while only using 60% parameters and 50% FLOPs. The code will be available at https://github.com/THU-Kingmin/IRRM.

Constrained text generation remains a challenging task, particularly when dealing with hard constraints. Traditional Natural Language Processing (NLP) approaches prioritize generating meaningful and coherent output. Also, the current state-of-the-art methods often lack the expressiveness and constraint satisfaction capabilities to handle such tasks effectively. This paper presents the Constraints First Framework to remedy this issue. This framework considers a constrained text generation problem as a discrete combinatorial optimization problem. It is solved by a constraint programming method that combines linguistic properties (e.g., n-grams or language level) with other more classical constraints (e.g., the number of characters, syllables, or words). Eventually, a curation phase allows for selecting the best-generated sentences according to perplexity using a large language model. The effectiveness of this approach is demonstrated by tackling a new more tediously constrained text generation problem: the iconic RADNER sentences problem. This problem aims to generate sentences respecting a set of quite strict rules defined by their use in vision and clinical research. Thanks to our CP-based approach, many new strongly constrained sentences have been successfully generated in an automatic manner. This highlights the potential of our approach to handle unreasonably constrained text generation scenarios.

Conventional machine learning methods are predominantly designed to predict outcomes based on a single data type. However, practical applications may encompass data of diverse types, such as text, images, and audio. We introduce interpretable tensor fusion (InTense), a multimodal learning method training a neural network to simultaneously learn multiple data representations and their interpretable fusion. InTense can separately capture both linear combinations and multiplicative interactions of the data types, thereby disentangling higher-order interactions from the individual effects of each modality. InTense provides interpretability out of the box by assigning relevance scores to modalities and their associations, respectively. The approach is theoretically grounded and yields meaningful relevance scores on multiple synthetic and real-world datasets. Experiments on four real-world datasets show that InTense outperforms existing state-of-the-art multimodal interpretable approaches in terms of accuracy and interpretability.

Transparency and explainability in image classification are essential for establishing trust in machine learning models and detecting biases and errors. State-of-the-art explainability methods generate saliency maps to show where a specific class is identified, without providing a detailed explanation of the model's decision process. Striving to address such a need, we introduce a post-hoc method that explains the entire feature extraction process of a Convolutional Neural Network. These explanations include a layer-wise representation of the features the model extracts from the input. Such features are represented as saliency maps generated by clustering and merging similar feature maps, to which we associate a weight derived by generalizing Grad-CAM for the proposed methodology. To further enhance these explanations, we include a set of textual labels collected through a gamified crowdsourcing activity and processed using NLP techniques and Sentence-BERT. Finally, we show an approach to generate global explanations by aggregating labels across multiple images.