Search

The integration of hardware accelerators has significantly advanced the capabilities of modern recommendation systems, enabling the exploration of complex ranking paradigms previously deemed impractical. However, the GPU-based computational costs present substantial challenges. In this paper, we demonstrate our development of an efficiency-driven approach to explore these paradigms, moving beyond traditional reliance on native PyTorch modules. We address the specific challenges posed by ranking models' dependence on categorical features, which vary in length and complicate GPU utilization. We introduce Jagged Feature Interaction Kernels, a novel method designed to extract fine-grained insights from long categorical features through efficient handling of dynamically sized tensors. We further enhance the performance of attention mechanisms by integrating Jagged tensors with Flash Attention. Our novel Jagged Flash Attention achieves up to 9x speedup and 22x memory reduction compared to dense attention. Notably, it also outperforms dense flash attention, with up to 3x speedup and 53% more memory efficiency. In production models, we observe 10% QPS improvement and 18% memory savings, enabling us to scale our recommendation systems with longer features and more complex architectures., Comment: 3 pages, 2 figures

Composed Image Retrieval (CIR) is a challenging task that aims to retrieve the target image based on a multimodal query, i.e., a reference image and its corresponding modification text. While previous supervised or zero-shot learning paradigms all fail to strike a good trade-off between time-consuming annotation cost and retrieval performance, recent researchers introduced the task of few-shot CIR (FS-CIR) and proposed a textual inversion-based network based on pretrained CLIP model to realize it. Despite its promising performance, the approach suffers from two key limitations: insufficient multimodal query composition training and indiscriminative training triplet selection. To address these two limitations, in this work, we propose a novel two-stage pseudo triplet guided few-shot CIR scheme, dubbed PTG-FSCIR. In the first stage, we employ a masked training strategy and advanced image caption generator to construct pseudo triplets from pure image data to enable the model to acquire primary knowledge related to multimodal query composition. In the second stage, based on active learning, we design a pseudo modification text-based query-target distance metric to evaluate the challenging score for each unlabeled sample. Meanwhile, we propose a robust top range-based random sampling strategy according to the 3-$\sigma$ rule in statistics, to sample the challenging samples for fine-tuning the pretrained model. Notably, our scheme is plug-and-play and compatible with any existing supervised CIR models. We tested our scheme across three backbones on three public datasets (i.e., FashionIQ, CIRR, and Birds-to-Words), achieving maximum improvements of 26.4%, 25.5% and 21.6% respectively, demonstrating our scheme's effectiveness., Comment: 15 pages, 5 figures

In this report, we present our approach for the Natural Language Query track and Goal Step track of the Ego4D Episodic Memory Benchmark at CVPR 2024. Both challenges require the localization of actions within long video sequences using textual queries. To enhance localization accuracy, our method not only processes the temporal information of videos but also identifies fine-grained objects spatially within the frames. To this end, we introduce a novel approach, termed ObjectNLQ, which incorporates an object branch to augment the video representation with detailed object information, thereby improving grounding efficiency. ObjectNLQ achieves a mean R@1 of 23.15, ranking 2nd in the Natural Language Queries Challenge, and gains 33.00 in terms of the metric R@1, IoU=0.3, ranking 3rd in the Goal Step Challenge. Our code will be released at https://github.com/Yisen-Feng/ObjectNLQ., Comment: The solution for the Natural Language Query track and Goal Step track at CVPR EgoVis Workshop 2024

In this report, we present our champion solution for Ego4D EgoSchema Challenge in CVPR 2024. To deeply integrate the powerful egocentric captioning model and question reasoning model, we propose a novel Hierarchical Comprehension scheme for egocentric video Question Answering, named HCQA. It consists of three stages: Fine-grained Caption Generation, Context-driven Summarization, and Inference-guided Answering. Given a long-form video, HCQA captures local detailed visual information and global summarised visual information via Fine-grained Caption Generation and Context-driven Summarization, respectively. Then in Inference-guided Answering, HCQA utilizes this hierarchical information to reason and answer given question. On the EgoSchema blind test set, HCQA achieves 75% accuracy in answering over 5,000 human curated multiple-choice questions. Our code will be released at https://github.com/Hyu-Zhang/HCQA., Comment: The champion solution for Ego4D EgoSchema Challenge in CVPR EgoVis Workshop 2024

The Visible-Infrared Person Re-identification (VI ReID) aims to match visible and infrared images of the same pedestrians across non-overlapped camera views. These two input modalities contain both invariant information, such as shape, and modality-specific details, such as color. An ideal model should utilize valuable information from both modalities during training for enhanced representational capability. However, the gap caused by modality-specific information poses substantial challenges for the VI ReID model to handle distinct modality inputs simultaneously. To address this, we introduce the Modality-aware and Instance-aware Visual Prompts (MIP) network in our work, designed to effectively utilize both invariant and specific information for identification. Specifically, our MIP model is built on the transformer architecture. In this model, we have designed a series of modality-specific prompts, which could enable our model to adapt to and make use of the specific information inherent in different modality inputs, thereby reducing the interference caused by the modality gap and achieving better identification. Besides, we also employ each pedestrian feature to construct a group of instance-specific prompts. These customized prompts are responsible for guiding our model to adapt to each pedestrian instance dynamically, thereby capturing identity-level discriminative clues for identification. Through extensive experiments on SYSU-MM01 and RegDB datasets, the effectiveness of both our designed modules is evaluated. Additionally, our proposed MIP performs better than most state-of-the-art methods., Comment: Accepyed by ACM International Conference on Multimedia Retrieval (ICMR'24)

The segmentation foundation model, e.g., Segment Anything Model (SAM), has attracted increasing interest in the medical image community. Early pioneering studies primarily concentrated on assessing and improving SAM's performance from the perspectives of overall accuracy and efficiency, yet little attention was given to the fairness considerations. This oversight raises questions about the potential for performance biases that could mirror those found in task-specific deep learning models like nnU-Net. In this paper, we explored the fairness dilemma concerning large segmentation foundation models. We prospectively curate a benchmark dataset of 3D MRI and CT scans of the organs including liver, kidney, spleen, lung and aorta from a total of 1056 healthy subjects with expert segmentations. Crucially, we document demographic details such as gender, age, and body mass index (BMI) for each subject to facilitate a nuanced fairness analysis. We test state-of-the-art foundation models for medical image segmentation, including the original SAM, medical SAM and SAT models, to evaluate segmentation efficacy across different demographic groups and identify disparities. Our comprehensive analysis, which accounts for various confounding factors, reveals significant fairness concerns within these foundational models. Moreover, our findings highlight not only disparities in overall segmentation metrics, such as the Dice Similarity Coefficient but also significant variations in the spatial distribution of segmentation errors, offering empirical evidence of the nuanced challenges in ensuring fairness in medical image segmentation., Comment: Accepted to MICCAI-2024

Accurate prediction of protein-ligand binding affinities is crucial for drug development. Recent advances in machine learning show promising results on this task. However, these methods typically rely heavily on labeled data, which can be scarce or unreliable, or they rely on assumptions like Boltzmann-distributed data that may not hold true in practice. Here, we present DualBind, a novel framework that integrates supervised mean squared error (MSE) with unsupervised denoising score matching (DSM) to accurately learn the binding energy function. DualBind not only addresses the limitations of DSM-only models by providing more accurate absolute affinity predictions but also improves generalizability and reduces reliance on labeled data compared to MSE-only models. Our experimental results demonstrate that DualBind excels in predicting binding affinities and can effectively utilize both labeled and unlabeled data to enhance performance., Comment: Preprint, work in progress

In recommendation systems, high-quality user embeddings can capture subtle preferences, enable precise similarity calculations, and adapt to changing preferences over time to maintain relevance. The effectiveness of recommendation systems depends on the quality of user embedding. We propose to asynchronously learn high fidelity user embeddings for billions of users each day from sequence based multimodal user activities through a Transformer-like large scale feature learning module. The async learned user representations embeddings (ALURE) are further converted to user similarity graphs through graph learning and then combined with user realtime activities to retrieval highly related ads candidates for the ads delivery system. Our method shows significant gains in both offline and online experiments., Comment: Accepted by workshop on Multimodal Representation and Retrieval at SIGIR 2024, Washington DC

White dwarfs represent the end stage for 97% of stars, making precise parameter measurement crucial for understanding stellar evolution. Traditional estimation methods involve fitting spectra or photometry, which require high-quality data. In recent years, machine learning has played a crucial role in processing spectral data due to its speed, automation, and accuracy. However, two common issues have been identified. First, most studies rely on data with high signal-to-noise ratios (SNR > 10), leaving many poor-quality datasets underutilized. Second, existing machine learning models, primarily based on convolutional networks, recurrent networks, and their variants, cannot simultaneously capture both the spatial and sequential information of spectra. To address these challenges, we designed the Estimator Network (EstNet), an advanced algorithm integrating multiple techniques, including Residual Networks, Squeeze and Excitation Attention, Gated Recurrent Units, Adaptive Loss, and Monte-Carlo Dropout Layers. We conducted parameter estimation on 5,965 poor-quality white dwarf spectra (R~1800, SNR~1.17), achieving average percentage errors of 14.86% for effective temperature and 3.97% for surface gravity. These results are significantly superior to other mainstream algorithms and consistent with the outcomes of traditional theoretical spectrum fitting methods. In the future, our algorithms will be applied for large-scale parameter estimation on the Chinese Space Station Telescope and the Large Synoptic Survey Telescope.

In recent years, with the rapid development of computer information technology, the development of artificial intelligence has been accelerating. The traditional geometry recognition technology is relatively backward and the recognition rate is low. In the face of massive information database, the traditional algorithm model inevitably has the problems of low recognition accuracy and poor performance. Deep learning theory has gradually become a very important part of machine learning. The implementation of convolutional neural network (CNN) reduces the difficulty of graphics generation algorithm. In this paper, using the advantages of lenet-5 architecture sharing weights and feature extraction and classification, the proposed geometric pattern recognition algorithm model is faster in the training data set. By constructing the shared feature parameters of the algorithm model, the cross-entropy loss function is used in the recognition process to improve the generalization of the model and improve the average recognition accuracy of the test data set., Comment: 6 pages,9 figures

Cardiac MRI, crucial for evaluating heart structure and function, faces limitations like slow imaging and motion artifacts. Undersampling reconstruction, especially data-driven algorithms, has emerged as a promising solution to accelerate scans and enhance imaging performance using highly under-sampled data. Nevertheless, the scarcity of publicly available cardiac k-space datasets and evaluation platform hinder the development of data-driven reconstruction algorithms. To address this issue, we organized the Cardiac MRI Reconstruction Challenge (CMRxRecon) in 2023, in collaboration with the 26th International Conference on MICCAI. CMRxRecon presented an extensive k-space dataset comprising cine and mapping raw data, accompanied by detailed annotations of cardiac anatomical structures. With overwhelming participation, the challenge attracted more than 285 teams and over 600 participants. Among them, 22 teams successfully submitted Docker containers for the testing phase, with 7 teams submitted for both cine and mapping tasks. All teams use deep learning based approaches, indicating that deep learning has predominately become a promising solution for the problem. The first-place winner of both tasks utilizes the E2E-VarNet architecture as backbones. In contrast, U-Net is still the most popular backbone for both multi-coil and single-coil reconstructions. This paper provides a comprehensive overview of the challenge design, presents a summary of the submitted results, reviews the employed methods, and offers an in-depth discussion that aims to inspire future advancements in cardiac MRI reconstruction models. The summary emphasizes the effective strategies observed in Cardiac MRI reconstruction, including backbone architecture, loss function, pre-processing techniques, physical modeling, and model complexity, thereby providing valuable insights for further developments in this field., Comment: 25 pages, 17 figures

Neural algorithmic reasoning is an emerging research direction that endows neural networks with the ability to mimic algorithmic executions step-by-step. A common paradigm in existing designs involves the use of historical embeddings in predicting the results of future execution steps. Our observation in this work is that such historical dependence intrinsically contradicts the Markov nature of algorithmic reasoning tasks. Based on this motivation, we present our ForgetNet, which does not use historical embeddings and thus is consistent with the Markov nature of the tasks. To address challenges in training ForgetNet at early stages, we further introduce G-ForgetNet, which uses a gating mechanism to allow for the selective integration of historical embeddings. Such an enhanced capability provides valuable computational pathways during the model's early training phase. Our extensive experiments, based on the CLRS-30 algorithmic reasoning benchmark, demonstrate that both ForgetNet and G-ForgetNet achieve better generalization capability than existing methods. Furthermore, we investigate the behavior of the gating mechanism, highlighting its degree of alignment with our intuitions and its effectiveness for robust performance., Comment: To appear at ICLR 2024 (Spotlight paper). 17 pages, 10 figures

The ever-expanding scale of integrated circuits has brought about a significant rise in the design risks associated with radiation-resistant integrated circuit chips. Traditional single-particle experimental methods, with their iterative design approach, are increasingly ill-suited for the challenges posed by large-scale integrated circuits. In response, this article introduces a novel sensitivity-aware single-particle radiation effects simulation framework tailored for System-on-Chip platforms. Based on SVM algorithm we have implemented fast finding and classification of sensitive circuit nodes. Additionally, the methodology automates soft error analysis across the entire software stack. The study includes practical experiments focusing on RISC-V architecture, encompassing core components, buses, and memory systems. It culminates in the establishment of databases for Single Event Upsets (SEU) and Single Event Transients (SET), showcasing the practical efficacy of the proposed methodology in addressing radiation-induced challenges at the scale of contemporary integrated circuits. Experimental results have shown up to 12.78X speed-up on the basis of achieving 94.58% accuracy., Comment: Accepted to the 61th ACM/IEEE Design Automation conference (DAC 2024)

Sarcasm Explanation in Dialogue (SED) is a new yet challenging task, which aims to generate a natural language explanation for the given sarcastic dialogue that involves multiple modalities (i.e., utterance, video, and audio). Although existing studies have achieved great success based on the generative pretrained language model BART, they overlook exploiting the sentiments residing in the utterance, video and audio, which are vital clues for sarcasm explanation. In fact, it is non-trivial to incorporate sentiments for boosting SED performance, due to three main challenges: 1) diverse effects of utterance tokens on sentiments; 2) gap between video-audio sentiment signals and the embedding space of BART; and 3) various relations among utterances, utterance sentiments, and video-audio sentiments. To tackle these challenges, we propose a novel sEntiment-enhanceD Graph-based multimodal sarcasm Explanation framework, named EDGE. In particular, we first propose a lexicon-guided utterance sentiment inference module, where a heuristic utterance sentiment refinement strategy is devised. We then develop a module named Joint Cross Attention-based Sentiment Inference (JCA-SI) by extending the multimodal sentiment analysis model JCA to derive the joint sentiment label for each video-audio clip. Thereafter, we devise a context-sentiment graph to comprehensively model the semantic relations among the utterances, utterance sentiments, and video-audio sentiments, to facilitate sarcasm explanation generation. Extensive experiments on the publicly released dataset WITS verify the superiority of our model over cutting-edge methods.

The age and mass of red giants are essential for understanding the structure and evolution of the Milky Way. Traditional isochrone methods for these estimations are inherently limited due to overlapping isochrones in the Hertzsprung-Russell diagram, while asteroseismology, though more precise, requires high-precision, long-term observations. In response to these challenges, we developed a novel framework, Spectral Transformer (SPT), to predict the age and mass of red giants aligned with asteroseismology from their spectra. A key component of SPT, the Multi-head Hadamard Self-Attention mechanism, designed specifically for spectra, can capture complex relationships across different wavelength. Further, we introduced a Mahalanobis distance-based loss function to address scale imbalance and interaction mode loss, and incorporated Monte Carlo dropout for quantitative analysis of prediction uncertainty.Trained and tested on 3,880 red giant spectra from LAMOST, the SPT achieved remarkable age and mass estimations with average percentage errors of 17.64% and 6.61%, respectively, and provided uncertainties for each corresponding prediction. The results significantly outperform those of traditional machine learning algorithms and demonstrate a high level of consistency with asteroseismology methods and isochrone fitting techniques. In the future, our work will leverage datasets from the Chinese Space Station Telescope and the Large Synoptic Survey Telescope to enhance the precision of the model and broaden its applicability in the field of astronomy and astrophysics., Comment: Accepted by A&A

Low-surface-brightness galaxies (LSBGs), fainter members of the galaxy population, are thought to be numerous. However, due to their low surface brightness, the search for a wide-area sample of LSBGs is difficult, which in turn limits our ability to fully understand the formation and evolution of galaxies as well as galaxy relationships. Edge-on LSBGs, due to their unique orientation, offer an excellent opportunity to study galaxy structure and galaxy components. In this work, we utilize the You Only Look Once object detection algorithm to construct an edge-on LSBG detection model by training on 281 edge-on LSBGs in Sloan Digital Sky Survey (SDSS) $gri$-band composite images. This model achieved a recall of 94.64% and a purity of 95.38% on the test set. We searched across 938,046 $gri$-band images from SDSS Data Release 16 and found 52,293 candidate LSBGs. To enhance the purity of the candidate LSBGs and reduce contamination, we employed the Deep Support Vector Data Description algorithm to identify anomalies within the candidate samples. Ultimately, we compiled a catalog containing 40,759 edge-on LSBG candidates. This sample has similar characteristics to the training data set, mainly composed of blue edge-on LSBG candidates. The catalog is available online at https://github.com/worldoutside/Edge-on_LSBG., Comment: 12 pages, 11 figures, accepted to be published on APJS

In contrast to conventional visual question answering, video-grounded dialog necessitates a profound understanding of both dialog history and video content for accurate response generation. Despite commendable progress made by existing approaches, they still face the challenges of incrementally understanding complex dialog history and assimilating video information. In response to these challenges, we present an iterative search and reasoning framework, which consists of a textual encoder, a visual encoder, and a generator. Specifically, we devise a path search and aggregation strategy in the textual encoder, mining core cues from dialog history that are pivotal to understanding the posed questions. Concurrently, our visual encoder harnesses an iterative reasoning network to extract and emphasize critical visual markers from videos, enhancing the depth of visual comprehension. Finally, we utilize the pre-trained GPT-2 model as our answer generator to decode the mined hidden clues into coherent and contextualized answers. Extensive experiments on three public datasets demonstrate the effectiveness and generalizability of our proposed framework.

In this paper, we present a novel problem, namely video timeline modeling. Our objective is to create a video-associated timeline from a set of videos related to a specific topic, thereby facilitating the content and structure understanding of the story being told. This problem has significant potential in various real-world applications, for instance, news story summarization. To bootstrap research in this area, we curate a realistic benchmark dataset, YouTube-News-Timeline, consisting of over $12$k timelines and $300$k YouTube news videos. Additionally, we propose a set of quantitative metrics to comprehensively evaluate and compare methodologies. With such a testbed, we further develop and benchmark several deep learning approaches to tackling this problem. We anticipate that this exploratory work will pave the way for further research in video timeline modeling. The assets are available via https://github.com/google-research/google-research/tree/master/video_timeline_modeling., Comment: Accepted as a spotlight by NeurIPS 2023, Track on Datasets and Benchmarks

Audiovisual data is everywhere in this digital age, which raises higher requirements for the deep learning models developed on them. To well handle the information of the multi-modal data is the key to a better audiovisual modal. We observe that these audiovisual data naturally have temporal attributes, such as the time information for each frame in the video. More concretely, such data is inherently multi-modal according to both audio and visual cues, which proceed in a strict chronological order. It indicates that temporal information is important in multi-modal acoustic event modeling for both intra- and inter-modal. However, existing methods deal with each modal feature independently and simply fuse them together, which neglects the mining of temporal relation and thus leads to sub-optimal performance. With this motivation, we propose a Temporal Multi-modal graph learning method for Acoustic event Classification, called TMac, by modeling such temporal information via graph learning techniques. In particular, we construct a temporal graph for each acoustic event, dividing its audio data and video data into multiple segments. Each segment can be considered as a node, and the temporal relationships between nodes can be considered as timestamps on their edges. In this case, we can smoothly capture the dynamic information in intra-modal and inter-modal. Several experiments are conducted to demonstrate TMac outperforms other SOTA models in performance. Our code is available at https://github.com/MGitHubL/TMac., Comment: This work has been accepted by ACM MM 2023 for publication

Cardiac magnetic resonance imaging (CMR) has emerged as a valuable diagnostic tool for cardiac diseases. However, a limitation of CMR is its slow imaging speed, which causes patient discomfort and introduces artifacts in the images. There has been growing interest in deep learning-based CMR imaging algorithms that can reconstruct high-quality images from highly under-sampled k-space data. However, the development of deep learning methods requires large training datasets, which have not been publicly available for CMR. To address this gap, we released a dataset that includes multi-contrast, multi-view, multi-slice and multi-coil CMR imaging data from 300 subjects. Imaging studies include cardiac cine and mapping sequences. Manual segmentations of the myocardium and chambers of all the subjects are also provided within the dataset. Scripts of state-of-the-art reconstruction algorithms were also provided as a point of reference. Our aim is to facilitate the advancement of state-of-the-art CMR image reconstruction by introducing standardized evaluation criteria and making the dataset freely accessible to the research community. Researchers can access the dataset at https://www.synapse.org/#!Synapse:syn51471091/wiki/., Comment: 14 pages, 8 figures

Sentiment analysis is a crucial task that aims to understand people's emotional states and predict emotional categories based on multimodal information. It consists of several subtasks, such as emotion recognition in conversation (ERC), aspect-based sentiment analysis (ABSA), and multimodal sentiment analysis (MSA). However, unifying all subtasks in sentiment analysis presents numerous challenges, including modality alignment, unified input/output forms, and dataset bias. To address these challenges, we propose a Task-Specific Prompt method to jointly model subtasks and introduce a multimodal generative framework called UniSA. Additionally, we organize the benchmark datasets of main subtasks into a new Sentiment Analysis Evaluation benchmark, SAEval. We design novel pre-training tasks and training methods to enable the model to learn generic sentiment knowledge among subtasks to improve the model's multimodal sentiment perception ability. Our experimental results show that UniSA performs comparably to the state-of-the-art on all subtasks and generalizes well to various subtasks in sentiment analysis., Comment: Accepted to ACM MM 2023