With the rapid evolution of the Internet of Things, many real-world applications utilize heterogeneously connected sensors to capture time-series information. Edge-based machine learning (ML) methodologies are often employed to analyze locally collected data. However, a fundamental issue across data-driven ML approaches is distribution shift. It occurs when a model is deployed on a data distribution different from what it was trained on, and can substantially degrade model performance. Additionally, increasingly sophisticated deep neural networks (DNNs) have been proposed to capture spatial and temporal dependencies in multi-sensor time series data, requiring intensive computational resources beyond the capacity of today's edge devices. While brain-inspired hyperdimensional computing (HDC) has been introduced as a lightweight solution for edge-based learning, existing HDCs are also vulnerable to the distribution shift challenge. In this paper, we propose DOMINO, a novel HDC learning framework addressing the distribution shift problem in noisy multi-sensor time-series data. DOMINO leverages efficient and parallel matrix operations on high-dimensional space to dynamically identify and filter out domain-variant dimensions. Our evaluation on a wide range of multi-sensor time series classification tasks shows that DOMINO achieves on average 2.04% higher accuracy than state-of-the-art (SOTA) DNN-based domain generalization techniques, and delivers 16.34x faster training and 2.89x faster inference. More importantly, DOMINO performs notably better when learning from partially labeled and highly imbalanced data, providing 10.93x higher robustness against hardware noises than SOTA DNNs.
Video Question Answering (VideoQA) is a challenging task that entails complex multi-modal reasoning. In contrast to multiple-choice VideoQA which aims to predict the answer given several options, the goal of open-ended VideoQA is to answer questions without restricting candidate answers. However, the majority of previous VideoQA models formulate open-ended VideoQA as a classification task to classify the video-question pairs into a fixed answer set, i.e., closed-vocabulary, which contains only frequent answers (e.g., top-1000 answers). This leads the model to be biased toward only frequent answers and fail to generalize on out-of-vocabulary answers. We hence propose a new benchmark, Open-vocabulary Video Question Answering (OVQA), to measure the generalizability of VideoQA models by considering rare and unseen answers. In addition, in order to improve the model's generalization power, we introduce a novel GNN-based soft verbalizer that enhances the prediction on rare and unseen answers by aggregating the information from their similar words. For evaluation, we introduce new baselines by modifying the existing (closed-vocabulary) open-ended VideoQA models and improve their performances by further taking into account rare and unseen answers. Our ablation studies and qualitative analyses demonstrate that our GNN-based soft verbalizer further improves the model performance, especially on rare and unseen answers. We hope that our benchmark OVQA can serve as a guide for evaluating the generalizability of VideoQA models and inspire future research. Code is available at https://github.com/mlvlab/OVQA.
Catastrophic forgetting; the loss of old knowledge upon acquiring new knowledge, is a pitfall faced by deep neural networks in real-world applications. Many prevailing solutions to this problem rely on storing exemplars (previously encountered data), which may not be feasible in applications with memory limitations or privacy constraints. Therefore, the recent focus has been on Non-Exemplar based Class Incremental Learning (NECIL) where a model incrementally learns about new classes without using any past exemplars. However, due to the lack of old data, NECIL methods struggle to discriminate between old and new classes causing their feature representations to overlap. We propose NAPA-VQ: Neighborhood Aware Prototype Augmentation with Vector Quantization, a framework that reduces this class overlap in NECIL. We draw inspiration from Neural Gas to learn the topological relationships in the feature space, identifying the neighboring classes that are most likely to get confused with each other. This neighborhood information is utilized to enforce strong separation between the neighboring classes as well as to generate old class representative prototypes that can better aid in obtaining a discriminative decision boundary between old and new classes. Our comprehensive experiments on CIFAR-100, TinyImageNet, and ImageNet-Subset demonstrate that NAPA-VQ outperforms the State-of-the-art NECIL methods by an average improvement of 5%, 2%, and 4% in accuracy and 10%, 3%, and 9% in forgetting respectively. Our code can be found in https://github.com/TamashaM/NAPA-VQ.git.
Data-driven algorithms for human-centered autonomy use observed data to compute models of human behavior in order to ensure safety, correctness, and to avoid potential errors that arise at runtime. However, such algorithms often neglect useful a priori knowledge, known as side information, that can improve the quality of data-driven models. We identify several key challenges in human-centered autonomy, and identify possible approaches to incorporate side information in data-driven models of human behavior.
In a joint vision-language space, a text feature (e.g., from "a photo of a dog") could effectively represent its relevant image features (e.g., from dog photos). Also, a recent study has demonstrated the cross-modal transferability phenomenon of this joint space. From these observations, we propose PromptStyler which simulates various distribution shifts in the joint space by synthesizing diverse styles via prompts without using any images to deal with source-free domain generalization. The proposed method learns to generate a variety of style features (from "a S* style of a") via learnable style word vectors for pseudo-words S*. To ensure that learned styles do not distort content information, we force style-content features (from "a S* style of a [class]") to be located nearby their corresponding content features (from "[class]") in the joint vision-language space. After learning style word vectors, we train a linear classifier using synthesized style-content features. PromptStyler achieves the state of the art on PACS, VLCS, OfficeHome and DomainNet, even though it does not require any images for training.
Personalized text generation is an emerging research area that has attracted much attention in recent years. Most studies in this direction focus on a particular domain by designing bespoke features or models. In this work, we propose a general approach for personalized text generation using large language models (LLMs). Inspired by the practice of writing education, we develop a multistage and multitask framework to teach LLMs for personalized generation. In writing instruction, the task of writing from sources is often decomposed into multiple steps that involve finding, evaluating, summarizing, synthesizing, and integrating information. Analogously, our approach to personalized text generation consists of multiple stages: retrieval, ranking, summarization, synthesis, and generation. In addition, we introduce a multitask setting that helps the model improve its generation ability further, which is inspired by the observation in education that a student's reading proficiency and writing ability are often correlated. We evaluate our approach on three public datasets, each of which covers a different and representative domain. Our results show significant improvements over a variety of baselines.
The delayed feedback problem is one of the most pressing challenges in predicting the conversion rate since users' conversions are always delayed in online commercial systems. Although new data are beneficial for continuous training, without complete feedback information, i.e., conversion labels, training algorithms may suffer from overwhelming fake negatives. Existing methods tend to use multitask learning or design data pipelines to solve the delayed feedback problem. However, these methods have a trade-off between data freshness and label accuracy. In this paper, we propose Delayed Feedback Modeling by Dynamic Graph Neural Network (DGDFEM). It includes three stages, i.e., preparing a data pipeline, building a dynamic graph, and training a CVR prediction model. In the model training, we propose a novel graph convolutional method named HLGCN, which leverages both high-pass and low-pass filters to deal with conversion and non-conversion relationships. The proposed method achieves both data freshness and label accuracy. We conduct extensive experiments on three industry datasets, which validate the consistent superiority of our method.
Multimodal Emotion Recognition in Conversation (ERC) has garnered growing attention from research communities in various fields. In this paper, we propose a cross-modal fusion network with emotion-shift awareness (CFN-ESA) for ERC. Extant approaches employ each modality equally without distinguishing the amount of emotional information, rendering it hard to adequately extract complementary and associative information from multimodal data. To cope with this problem, in CFN-ESA, textual modalities are treated as the primary source of emotional information, while visual and acoustic modalities are taken as the secondary sources. Besides, most multimodal ERC models ignore emotion-shift information and overfocus on contextual information, leading to the failure of emotion recognition under emotion-shift scenario. We elaborate an emotion-shift module to address this challenge. CFN-ESA mainly consists of the unimodal encoder (RUME), cross-modal encoder (ACME), and emotion-shift module (LESM). RUME is applied to extract conversation-level contextual emotional cues while pulling together the data distributions between modalities; ACME is utilized to perform multimodal interaction centered on textual modality; LESM is used to model emotion shift and capture related information, thereby guide the learning of the main task. Experimental results demonstrate that CFN-ESA can effectively promote performance for ERC and remarkably outperform the state-of-the-art models.
In recent years, 3D representation learning has turned to 2D vision-language pre-trained models to overcome data scarcity challenges. However, existing methods simply transfer 2D alignment strategies, aligning 3D representations with single-view 2D images and coarse-grained parent category text. These approaches introduce information degradation and insufficient synergy issues, leading to performance loss. Information degradation arises from overlooking the fact that a 3D representation should be equivalent to a series of multi-view images and more fine-grained subcategory text. Insufficient synergy neglects the idea that a robust 3D representation should align with the joint vision-language space, rather than independently aligning with each modality. In this paper, we propose a multi-view joint modality modeling approach, termed JM3D, to obtain a unified representation for point cloud, text, and image. Specifically, a novel Structured Multimodal Organizer (SMO) is proposed to address the information degradation issue, which introduces contiguous multi-view images and hierarchical text to enrich the representation of vision and language modalities. A Joint Multi-modal Alignment (JMA) is designed to tackle the insufficient synergy problem, which models the joint modality by incorporating language knowledge into the visual modality. Extensive experiments on ModelNet40 and ScanObjectNN demonstrate the effectiveness of our proposed method, JM3D, which achieves state-of-the-art performance in zero-shot 3D classification. JM3D outperforms ULIP by approximately 4.3% on PointMLP and achieves an improvement of up to 6.5% accuracy on PointNet++ in top-1 accuracy for zero-shot 3D classification on ModelNet40. The source code and trained models for all our experiments are publicly available at https://github.com/Mr-Neko/JM3D.
We can estimate the size of the speaker solely based on their speech sounds. We had proposed an auditory computational theory of the stabilised wavelet-Mellin transform (SWMT), which segregates information about the size and shape of vocal tract and glottal vibration, to explain this observation. It was demonstrated that the auditory representation or excitation pattern (EP) associated with a weighting function based on SWMT, referred to as "SSI weigh", made it possible to explain the psychometric functions of size perception. In this study, we investigated whether EP with SSI weight can precisely estimate vocal tract lengths (VTLs) which were measured using male and female MRI data. It was found that the use of SSI weight significantly improved the VTL estimation. Moreover, the estimation errors were significantly smaller in the EP with the SSI weight than those in the commonly used spectra derived from the Fourier transform, Mel filterbank, and WORLD vocoder. It was also shown that the SSI weight can be easily introduced into these spectra to improve the performance.