AAAI 2024 Papers

Previous 3D hand pose estimation methods primarily rely on a single modality, either RGB or depth, and the comprehensive utilization of the dual modalities has not been extensively explored. RGB and depth data provide complementary information and thus can be fused to enhance the robustness of 3D hand pose estimation. However, there exist two problems for applying existing fusion methods in 3D hand pose estimation: redundancy of dense feature fusion and ambiguity of visual features. First, pixel-wise feature interactions introduce high computational costs and ineffective calculations of invalid pixels. Second, visual features suffer from ambiguity due to color and texture similarities, as well as depth holes and noise caused by frequent hand movements, which interferes with modeling cross-modal correlations. In this paper, we propose Keypoint-Fusion for RGB-D based 3D hand pose estimation, which leverages the unique advantages of dual modalities to mutually eliminate the feature ambiguity, and performs cross-modal feature fusion in a more efficient way. Specifically, we focus cross-modal fusion on sparse yet informative spatial regions (i.e. keypoints). Meanwhile, by explicitly extracting relatively more reliable information as disambiguation evidence, depth modality provides 3D geometric information for RGB feature pixels, and RGB modality complements the precise edge information lost due to the depth noise. Keypoint-Fusion achieves state-of-the-art performance on two challenging hand datasets, significantly decreasing the error compared with previous single-modal methods.

In this paper, we study the problem of estimating the normalizing constant through queries to the black-box function f, which is the integration of the exponential function of f scaled by a problem parameter lambda. We assume f belongs to a reproducing kernel Hilbert space (RKHS), and show that to estimate the normalizing constant within a small relative error, the level of difficulty depends on the value of lambda: When lambda approaches zero, the problem is similar to Bayesian quadrature (BQ), while when lambda approaches infinity, the problem is similar to Bayesian optimization (BO). More generally, the problem varies between BQ and BO. We find that this pattern holds true even when the function evaluations are noisy, bringing new aspects to this topic. Our findings are supported by both algorithm-independent lower bounds and algorithmic upper bounds, as well as simulation studies conducted on a variety of benchmark functions.

An exoplanet is a planet, which is not a part of our solar system. Whether life exists in one or more of these exoplanets has fascinated humans for centuries. NASA’s Kepler Space Telescope has discovered more than 70% of known exoplanets in our universe. However, manually determining whether a Kepler light curve indicates an exoplanet or not becomes infeasible with the large volume of data. Due to this, we propose a deep learning-based strategy to automatically classify a Kepler light curve. More specifically, we first convert the light curve time series into its corresponding Markov Transition Field (MTF) image and then classify it. Results show that the accuracy of the proposed technique is 99.39%, which is higher than all current state-of-the-art approaches.

Explaining predictions of black-box neural networks is crucial when applied to decision-critical tasks. Thus, attribution maps are commonly used to identify important image regions, despite prior work showing that humans prefer explanations based on similar examples. To this end, ProtoPNet learns a set of class-representative feature vectors (prototypes) for case-based reasoning. During inference, similarities of latent features to prototypes are linearly classified to form predictions and attribution maps are provided to explain the similarity. In this work, we evaluate whether architectures for case-based reasoning fulfill established axioms required for faithful explanations using the example of ProtoPNet. We show that such architectures allow the extraction of faithful explanations. However, we prove that the attribution maps used to explain the similarities violate the axioms. We propose a new procedure to extract explanations for trained ProtoPNets, named ProtoPFaith. Conceptually, these explanations are Shapley values, calculated on the similarity scores of each prototype. They allow to faithfully answer which prototypes are present in an unseen image and quantify each pixel’s contribution to that presence, thereby complying with all axioms. The theoretical violations of ProtoPNet manifest in our experiments on three datasets (CUB-200-2011, Stanford Dogs, RSNA) and five architectures (ConvNet, ResNet, ResNet50, WideResNet50, ResNeXt50). Our experiments show a qualitative difference between the explanations given by ProtoPNet and ProtoPFaith. Additionally, we quantify the explanations with the Area Over the Perturbation Curve, on which ProtoPFaith outperforms ProtoPNet on all experiments by a factor >10^3.

Dual-lens super-resolution (SR) is a practical scenario for reference (Ref) based SR by utilizing the telephoto image (Ref) to assist the super-resolution of the low-resolution wide-angle image (LR input). Different from general RefSR, the Ref in dual-lens SR only covers the overlapped field of view (FoV) area. However, current dual-lens SR methods rarely utilize these specific characteristics and directly perform dense matching between the LR input and Ref. Due to the resolution gap between LR and Ref, the matching may miss the best-matched candidate and destroy the consistent structures in the overlapped FoV area. Different from them, we propose to first align the Ref with the center region (namely the overlapped FoV area) of the LR input by combining global warping and local warping to make the aligned Ref be sharp and consistent. Then, we formulate the aligned Ref and LR center as value-key pairs, and the corner region of the LR is formulated as queries. In this way, we propose a kernel-free matching strategy by matching between the LR-corner (query) and LR-center (key) regions, and the corresponding aligned Ref (value) can be warped to the corner region of the target. Our kernel-free matching strategy avoids the resolution gap between LR and Ref, which makes our network have better generalization ability. In addition, we construct a DuSR-Real dataset with (LR, Ref, HR) triples, where the LR and HR are well aligned. Experiments on three datasets demonstrate that our method outperforms the second-best method by a large margin. Our code and dataset are available at https://github.com/ZifanCui/KeDuSR.

Our life is populated with articulated objects. Current category-level articulation estimation works largely focus on predicting part-level 6D poses on static point cloud observations. In this paper, we tackle the problem of category-level online robust and real-time 6D pose tracking of articulated objects, where we propose KPA-Tracker, a novel 3D KeyPoint based Articulated object pose Tracker. Given an RGB-D image or a partial point cloud at the current frame as well as the estimated per-part 6D poses from the last frame, our KPA-Tracker can effectively update the poses with learned 3D keypoints between the adjacent frames. Specifically, we first canonicalize the input point cloud and formulate the pose tracking as an inter-frame pose increment estimation task. To learn consistent and separate 3D keypoints for every rigid part, we build KPA-Gen that outputs the high-quality ordered 3D keypoints in an unsupervised manner. During pose tracking on the whole video, we further propose a keypoint-based articulation tracking algorithm that mines keyframes as reference for accurate pose updating. We provide extensive experiments on validating our KPA-Tracker on various datasets ranging from synthetic point cloud observation to real-world scenarios, which demonstrates the superior performance and robustness of the KPA-Tracker. We believe that our work has the potential to be applied in many fields including robotics, embodied intelligence and augmented reality. All the datasets and codes are available at https://github.com/hhhhhar/KPA-Tracker.

Trajectory similarity computation serves as a fundamental functionality of various spatial information applications. Although existing deep learning similarity computation methods offer better efficiency and accuracy than non-learning solutions, they are still immature in trajectory embedding and suffer from poor generality and heavy preprocessing for training. Targeting these limitations, we propose a novel framework named KGTS based on knowledge graph grid embedding, prompt trajectory embedding, and unsupervised contrastive learning for improved trajectory similarity computation. Specifically, we first embed map grids with a GRot embedding method to vigorously grasp the neighbouring relations of grids. Then, a prompt trajectory embedding network incorporates the resulting grid embedding and extracts trajectory structure and point order information. It is trained by unsupervised contrastive learning, which not only alleviates the heavy preprocessing burden but also provides exceptional generality with creatively designed strategies for positive sample generation. The prompt trajectory embedding adopts a customized prompt paradigm to mitigate the gap between the grid embedding and the trajectory embedding. Extensive experiments on two real-world trajectory datasets demonstrate the superior performance of KGTS over state-of-the-art methods.

Knowledge graph embedding (KGE) is an efficient and scalable method for knowledge graph completion. However, most existing KGE methods suffer from the challenge of multiple relation semantics, which often degrades their performance. This is because most KGE methods learn fixed continuous vectors for entities (relations) and make deterministic entity predictions to complete the knowledge graph, which hardly captures multiple relation semantics. To tackle this issue, previous works try to learn complex probabilistic embeddings instead of fixed embeddings but suffer from heavy computational complexity. In contrast, this paper proposes a simple yet efficient framework namely the Knowledge Graph Diffusion Model (KGDM) to capture the multiple relation semantics in prediction. Its key idea is to cast the problem of entity prediction into conditional entity generation. Specifically, KGDM estimates the probabilistic distribution of target entities in prediction through Denoising Diffusion Probabilistic Models (DDPM). To bridge the gap between continuous diffusion models and discrete KGs, two learnable embedding functions are defined to map entities and relation to continuous vectors. To consider connectivity patterns of KGs, a Conditional Entity Denoiser model is introduced to generate target entities conditioned on given entities and relations. Extensive experiments demonstrate that KGDM significantly outperforms existing state-of-the-art methods in three benchmark datasets.

Treatment effect estimation (TEE) is the task of determining the impact of various treatments on patient outcomes. Current TEE methods fall short due to reliance on limited labeled data and challenges posed by sparse and high-dimensional observational patient data. To address the challenges, we introduce a novel pre-training and fine-tuning framework, KG-TREAT, which synergizes large-scale observational patient data with biomedical knowledge graphs (KGs) to enhance TEE. Unlike previous approaches, KG-TREAT constructs dual-focus KGs and integrates a deep bi-level attention synergy method for in-depth information fusion, enabling distinct encoding of treatment-covariate and outcome-covariate relationships. KG-TREAT also incorporates two pre-training tasks to ensure a thorough grounding and contextualization of patient data and KGs. Evaluation on four downstream TEE tasks shows KG-TREAT’s superiority over existing methods, with an average improvement of 7% in Area under the ROC Curve (AUC) and 9% in Influence Function-based Precision of Estimating Heterogeneous Effects (IF-PEHE). The effectiveness of our estimated treatment effects is further affirmed by alignment with established randomized clinical trial findings.

The Maximum k-Defective Clique Problem (MDCP) aims to find a maximum k-defective clique in a given graph, where a k-defective clique is a relaxation clique missing at most k edges. MDCP is NP-hard and finds many real-world applications in analyzing dense but not necessarily complete subgraphs. Exact algorithms for MDCP mainly follow the Branch-and-bound (BnB) framework, whose performance heavily depends on the quality of the upper bound on the cardinality of a maximum k-defective clique. The state-of-the-art BnB MDCP algorithms calculate the upper bound quickly but conservatively as they ignore many possible missing edges. In this paper, we propose a novel CoLoring-based Upper Bound (CLUB) that uses graph coloring techniques to detect independent sets so as to detect missing edges ignored by the previous methods. We then develop a new BnB algorithm for MDCP, called KD-Club, using CLUB in both the preprocessing stage for graph reduction and the BnB searching process for branch pruning. Extensive experiments show that KD-Club significantly outperforms state-of-the-art BnB MDCP algorithms on the number of solved instances within the cut-off time, having much smaller search tree and shorter solving time on various benchmarks.

Automating the processing of health insurance claims to achieve "Straight-Through Processing" is one of the holy grails that all insurance companies aim to achieve. One of the major impediments to this automation is the difficulty in establishing the relationship between the underwriting exclusions that a policy has and the incoming claim's diagnosis information. Typically, policy underwriting exclusions are captured in free-text such as "Respiratory illnesses are excluded due to a pre-existing asthma condition". A medical claim coming from a hospital would have the diagnosis represented using the International Classification of Disease (ICD) codes from the World Health Organization. The complex and labour-intensive task of establishing the relationship between free-text underwriting exclusions in health insurance policies and medical diagnosis codes from health insurance claims is critical towards determining if a claim should be rejected due to underwriting exclusions. In this work, we present a novel framework that leverages both explicit and implicit domain knowledge present in medical ontologies and pre-trained language models respectively, to effectively establish the relationship between free-text describing medical conditions present in underwriting exclusions and the ICD-10CM diagnosis codes in health insurance claims. Termed KAMEL (Knowledge Aware Medical Entity Linkage), our proposed framework addresses the limitations faced by prior approaches when evaluated on real-world health insurance claims data. Our proposed framework have been deployed in several multi-national health insurance providers to automate their health insurance claims.

Large Language Models (LLMs) have demonstrated impressive performance in natural language processing tasks by leveraging chain of thought (CoT) that enables step-by-step thinking. Extending LLMs with multimodal capabilities is the recent interest, but incurs computational cost and requires substantial hardware resources. To address these challenges, we propose KAM-CoT a framework that integrates CoT reasoning, Knowledge Graphs (KGs), and multiple modalities for a comprehensive understanding of multimodal tasks. KAM-CoT adopts a two-stage training process with KG grounding to generate effective rationales and answers. By incorporating external knowledge from KGs during reasoning, the model gains a deeper contextual understanding reducing hallucinations and enhancing the quality of answers. This knowledge-augmented CoT reasoning empowers the model to handle questions requiring external context, providing more informed answers. Experimental findings show KAM-CoT outperforms the state-of-the-art methods. On the ScienceQA dataset, we achieve an average accuracy of 93.87%, surpassing GPT-3.5 (75.17%) by 18% and GPT-4 (83.99%) by 10%. Remarkably, KAM-CoT achieves these results with only 280M trainable parameters at a time, demonstrating its cost-efficiency and effectiveness.

Pre-trained language models (PLMs) contain vast amounts of factual knowledge, but how the knowledge is stored in the parameters remains unclear. This paper delves into the complex task of understanding how factual knowledge is stored in multilingual PLMs, and introduces the Architecture-adapted Multilingual Integrated Gradients method, which successfully localizes knowledge neurons more precisely compared to current methods, and is more universal across various architectures and languages. Moreover, we conduct an in-depth exploration on knowledge neurons, leading to the following two important discoveries: (1) The discovery of Language-Independent Knowledge Neurons, which store factual knowledge in a form that transcends language. We design cross-lingual knowledge editing experiments, demonstrating that the PLMs can accomplish this task based on language-independent neurons; (2) The discovery of Degenerate Knowledge Neurons, a novel type of neuron showing that different knowledge neurons can store the same fact. Its property of functional overlap endows the PLMs with a robust mastery of factual knowledge. We design fact-checking experiments, proving that the degenerate knowledge neurons can help the PLMs to detect wrong facts. Experiments corroborate these findings, shedding light on the mechanisms of factual knowledge storage in multilingual PLMs, and contribute valuable insights to the field. The code is available at https://github.com/heng840/AMIG.

Tactile signals collected by wearable electronics are essential in modeling and understanding human behavior. One of the main applications of tactile signals is action classification, especially in healthcare and robotics. However, existing tactile classification methods fail to capture the spatial and temporal features of tactile signals simultaneously, which results in sub-optimal performances. In this paper, we design Spatio-Temporal Aware tactility Transformer (STAT) to utilize continuous tactile signals for action classification. We propose spatial and temporal embeddings along with a new temporal pretraining task in our model, which aims to enhance the transformer in modeling the spatio-temporal features of tactile signals. Specially, the designed temporal pretraining task is to differentiate the time order of tubelet inputs to model the temporal properties explicitly. Experimental results on a public action classification dataset demonstrate that our model outperforms state-of-the-art methods in all metrics.

We propose a novel single-loop decentralized algorithm, DGDA-VR, for solving the stochastic nonconvex strongly-concave minimax problems over a connected network of agents, which are equipped with stochastic first-order oracles to estimate their local gradients. DGDA-VR, incorporating variance reduction, achieves O(ε^−3) oracle complexity and O(ε^−2) communication complexity without resorting to multi-communication rounds – both are optimal, i.e., matching the lower bounds for this class of problems. Since DGDA-VR does not require multiple communication rounds, it is applicable to a broader range of decentralized computational environments. To the best of our knowledge, this is the first distributed method using a single communication round in each iteration to jointly optimize the oracle and communication complexities for the problem considered here.

Determining the types of neurons within a nervous system plays a significant role in the analysis of brain connectomics and the investigation of neurological diseases. However, the efficiency of utilizing anatomical, physiological, or molecular characteristics of neurons is relatively low and costly. With the advancements in electron microscopy imaging and analysis techniques for brain tissue, we are able to obtain whole-brain connectome consisting neuronal high-resolution morphology and connectivity information. However, few models are built based on such data for automated neuron classification. In this paper, we propose NeuNet, a framework that combines morphological information of neurons obtained from skeleton and topological information between neurons obtained from neural circuit. Specifically, NeuNet consists of three components, namely Skeleton Encoder, Connectome Encoder, and Readout Layer. Skeleton Encoder integrates the local information of neurons in a bottom-up manner, with a one-dimensional convolution in neural skeleton's point data; Connectome Encoder uses a graph neural network to capture the topological information of neural circuit; finally, Readout Layer fuses the above two information and outputs classification results. We reprocess and release two new datasets for neuron classification task from volume electron microscopy(VEM) images of human brain cortex and Drosophila brain. Experiments on these two datasets demonstrated the effectiveness of our model with accuracies of 0.9169 and 0.9363, respectively. Code and data are available at: https://github.com/WHUminghui/NeuNet.

As a neuromorphic camera with high temporal resolution, spike camera can capture dynamic scenes with high-speed motion. Recently, spike camera with a color filter array (CFA) has been developed for color imaging. There are some methods for spike camera demosaicing to reconstruct color images from Bayer-pattern spike streams. However, the demosaicing results are bothered by severe noise in spike streams, to which previous works pay less attention. In this paper, we propose an iterative joint demosaicing and denoising network (SJDD-Net) for spike cameras based on the observation model. Firstly, we design a color spike representation (CSR) to learn latent representation from Bayer-pattern spike streams. In CSR, we propose an offset-sharing deformable convolution module to align temporal features of color channels. Then we develop a spike noise estimator (SNE) to obtain features of the noise distribution. Finally, a color correlation prior (CCP) module is proposed to utilize the color correlation for better details. For training and evaluation, we designed a spike camera simulator to generate Bayer-pattern spike streams with synthesized noise. Besides, we captured some Bayer-pattern spike streams, building the first real-world captured dataset to our knowledge. Experimental results show that our method can restore clean images from Bayer-pattern spike streams. The source codes and dataset are available at https://github.com/csycdong/SJDD-Net.

Time-series and text data are prevalent in healthcare and frequently co-exist, yet they are typically modeled in isolation. Even studies that jointly model time-series and text, do so by converting time-series to images or graphs. We hypothesize that explicitly modeling time-series jointly with text can improve tasks such as summarization and question answering for time-series data, which have received little attention so far. To address this gap, we introduce JoLT to jointly learn desired representations from pre-trained time-series and text models. JoLT utilizes a Querying Transformer (Q-Former) to align the time-series and text representations. Our experiments on a large real-world electrocardiography dataset for medical time-series summarization show that JoLT outperforms state-of-the-art image captioning approaches.

Real-world image super-resolution (RWSR) is a long-standing problem as low-quality (LQ) images often have complex and unidentified degradations. Existing methods such as Generative Adversarial Networks (GANs) or continuous diffusion models present their own issues including GANs being difficult to train while continuous diffusion models requiring numerous inference steps. In this paper, we propose an Iterative Token Evaluation and Refinement (ITER) framework for RWSR, which utilizes a discrete diffusion model operating in the discrete token representation space, i.e., indexes of features extracted from a VQGAN codebook pre-trained with high-quality (HQ) images. We show that ITER is easier to train than GANs and more efficient than continuous diffusion models. Specifically, we divide RWSR into two sub-tasks, i.e., distortion removal and texture generation. Distortion removal involves simple HQ token prediction with LQ images, while texture generation uses a discrete diffusion model to iteratively refine the distortion removal output with a token refinement network. In particular, we propose to include a token evaluation network in the discrete diffusion process. It learns to evaluate which tokens are good restorations and helps to improve the iterative refinement results. Moreover, the evaluation network can first check status of the distortion removal output and then adaptively select total refinement steps needed, thereby maintaining a good balance between distortion removal and texture generation. Extensive experimental results show that ITER is easy to train and performs well within just 8 iterative steps.