Deep learning models for semantic segmentation often experience performance degradation when deployed to unseen target domains unidentified during the training phase. This is mainly due to variations in image texture (\ie style) from different data sources. To tackle this challenge, existing domain generalized semantic segmentation (DGSS) methods attempt to remove style variations from the feature. However, these approaches struggle with the entanglement of style and content, which may lead to the unintentional removal of crucial content information, causing performance degradation. This study addresses this limitation by proposing BlindNet, a novel DGSS approach that blinds the style without external modules or datasets. The main idea behind our proposed approach is to alleviate the effect of style in the encoder whilst facilitating robust segmentation in the decoder. To achieve this, BlindNet comprises two key components: covariance alignment and semantic consistency contrastive learning. Specifically, the covariance alignment trains the encoder to uniformly recognize various styles and preserve the content information of the feature, rather than removing the style-sensitive factor. Meanwhile, semantic consistency contrastive learning enables the decoder to construct discriminative class embedding space and disentangles features that are vulnerable to misclassification. Through extensive experiments, our approach outperforms existing DGSS methods, exhibiting robustness and superior performance for semantic segmentation on unseen target domains.
We present Shape Non-rigid Kinematics (SNK), a novel zero-shot method for non-rigid shape matching that eliminates the need for extensive training or ground truth data. SNK operates on a single pair of shapes, and employs a reconstruction-based strategy using an encoder-decoder architecture, which deforms the source shape to closely match the target shape. During the process, an unsupervised functional map is predicted and converted into a point-to-point map, serving as a supervisory mechanism for the reconstruction. To aid in training, we have designed a new decoder architecture that generates smooth, realistic deformations. SNK demonstrates competitive results on traditional benchmarks, simplifying the shape-matching process without compromising accuracy. Our code can be found online: https://github.com/pvnieo/SNK
Large Language models (LLMs) have demonstrated impressive in-context learning (ICL) capabilities, where a LLM makes predictions for a given test input together with a few input-output pairs (demonstrations). Nevertheless, the inclusion of demonstrations leads to a quadratic increase in the computational overhead of the self-attention mechanism. Existing solutions attempt to distill lengthy demonstrations into compact vectors. However, they often require task-specific retraining or compromise LLM's in-context learning performance. To mitigate these challenges, we present Meta dEmonstratioN Distillation (MEND), where a language model learns to distill any lengthy demonstrations into vectors without retraining for a new downstream task. We exploit the knowledge distillation to enhance alignment between MEND and LLM, achieving both efficiency and effectiveness simultaneously. MEND is endowed with the meta-knowledge of distilling demonstrations through a two-stage training process, which includes meta-distillation pretraining and fine-tuning. Comprehensive evaluations across seven diverse ICL task partitions using decoder-only (GPT-2) and encoder-decoder (T5) attest to MEND's prowess. It not only matches but often outperforms the Vanilla ICL as well as other state-of-the-art distillation models, while significantly reducing the computational demands. This innovation promises enhanced scalability and efficiency for the practical deployment of large language models
Studies have shown that in noisy acoustic environments, providing binaural signals to the user of an assistive listening device may improve speech intelligibility and spatial awareness. This paper presents a binaural speech enhancement method using a complex convolutional neural network with an encoder-decoder architecture and a complex multi-head attention transformer. The model is trained to estimate individual complex ratio masks in the time-frequency domain for the left and right-ear channels of binaural hearing devices. The model is trained using a novel loss function that incorporates the preservation of spatial information along with speech intelligibility improvement and noise reduction. Simulation results for acoustic scenarios with a single target speaker and isotropic noise of various types show that the proposed method improves the estimated binaural speech intelligibility and preserves the binaural cues better in comparison with several baseline algorithms.
Recently, decoder-only pre-trained large language models (LLMs), with several tens of billion parameters, have significantly impacted a wide range of natural language processing (NLP) tasks. While encoder-only or encoder-decoder pre-trained language models have already proved to be effective in discourse parsing, the extent to which LLMs can perform this task remains an open research question. Therefore, this paper explores how beneficial such LLMs are for Rhetorical Structure Theory (RST) discourse parsing. Here, the parsing process for both fundamental top-down and bottom-up strategies is converted into prompts, which LLMs can work with. We employ Llama 2 and fine-tune it with QLoRA, which has fewer parameters that can be tuned. Experimental results on three benchmark datasets, RST-DT, Instr-DT, and the GUM corpus, demonstrate that Llama 2 with 70 billion parameters in the bottom-up strategy obtained state-of-the-art (SOTA) results with significant differences. Furthermore, our parsers demonstrated generalizability when evaluated on RST-DT, showing that, in spite of being trained with the GUM corpus, it obtained similar performances to those of existing parsers trained with RST-DT.
We present JointMotion, a self-supervised pre-training method for joint motion prediction in self-driving vehicles. Our method jointly optimizes a scene-level objective connecting motion and environments, and an instance-level objective to refine learned representations. Scene-level representations are learned via non-contrastive similarity learning of past motion sequences and environment context. At the instance level, we use masked autoencoding to refine multimodal polyline representations. We complement this with an adaptive pre-training decoder that enables JointMotion to generalize across different environment representations, fusion mechanisms, and dataset characteristics. Notably, our method reduces the joint final displacement error of Wayformer, HPTR, and Scene Transformer models by 3\%, 8\%, and 12\%, respectively; and enables transfer learning between the Waymo Open Motion and the Argoverse 2 Motion Forecasting datasets. Code: https://github.com/kit-mrt/future-motion
We propose a novel concept of dual and integrated latent topologies (DITTO in short) for implicit 3D reconstruction from noisy and sparse point clouds. Most existing methods predominantly focus on single latent type, such as point or grid latents. In contrast, the proposed DITTO leverages both point and grid latents (i.e., dual latent) to enhance their strengths, the stability of grid latents and the detail-rich capability of point latents. Concretely, DITTO consists of dual latent encoder and integrated implicit decoder. In the dual latent encoder, a dual latent layer, which is the key module block composing the encoder, refines both latents in parallel, maintaining their distinct shapes and enabling recursive interaction. Notably, a newly proposed dynamic sparse point transformer within the dual latent layer effectively refines point latents. Then, the integrated implicit decoder systematically combines these refined latents, achieving high-fidelity 3D reconstruction and surpassing previous state-of-the-art methods on object- and scene-level datasets, especially in thin and detailed structures.
Variational Autoencoders for multimodal data hold promise for many tasks in data analysis, such as representation learning, conditional generation, and imputation. Current architectures either share the encoder output, decoder input, or both across modalities to learn a shared representation. Such architectures impose hard constraints on the model. In this work, we show that a better latent representation can be obtained by replacing these hard constraints with a soft constraint. We propose a new mixture-of-experts prior, softly guiding each modality's latent representation towards a shared aggregate posterior. This approach results in a superior latent representation and allows each encoding to preserve information better from its uncompressed original features. In extensive experiments on multiple benchmark datasets and two challenging real-world datasets, we show improved learned latent representations and imputation of missing data modalities compared to existing methods.
Deep learning models, particularly those based on transformers, often employ numerous stacked structures, which possess identical architectures and perform similar functions. While effective, this stacking paradigm leads to a substantial increase in the number of parameters, posing challenges for practical applications. In today's landscape of increasingly large models, stacking depth can even reach dozens, further exacerbating this issue. To mitigate this problem, we introduce LORS (LOw-rank Residual Structure). LORS allows stacked modules to share the majority of parameters, requiring a much smaller number of unique ones per module to match or even surpass the performance of using entirely distinct ones, thereby significantly reducing parameter usage. We validate our method by applying it to the stacked decoders of a query-based object detector, and conduct extensive experiments on the widely used MS COCO dataset. Experimental results demonstrate the effectiveness of our method, as even with a 70\% reduction in the parameters of the decoder, our method still enables the model to achieve comparable or
Prohibited item detection in X-ray images is one of the most essential and highly effective methods widely employed in various security inspection scenarios. Considering the significant overlapping phenomenon in X-ray prohibited item images, we propose an Anti-Overlapping DETR (AO-DETR) based on one of the state-of-the-art general object detectors, DINO. Specifically, to address the feature coupling issue caused by overlapping phenomena, we introduce the Category-Specific One-to-One Assignment (CSA) strategy to constrain category-specific object queries in predicting prohibited items of fixed categories, which can enhance their ability to extract features specific to prohibited items of a particular category from the overlapping foreground-background features. To address the edge blurring problem caused by overlapping phenomena, we propose the Look Forward Densely (LFD) scheme, which improves the localization accuracy of reference boxes in mid-to-high-level decoder layers and enhances the ability to locate blurry edges of the final layer. Similar to DINO, our AO-DETR provides two different versions with distinct backbones, tailored to meet diverse application requirements. Extensive experiments on the PIXray and OPIXray datasets demonstrate that the proposed method surpasses the state-of-the-art object detectors, indicating its potential applications in the field of prohibited item detection. The source code will be released at https://github.com/Limingyuan001/AO-DETR-test.
Video generation necessitates both global coherence and local realism. This work presents a novel non-autoregressive method GLOBER, which first generates global features to obtain comprehensive global guidance and then synthesizes video frames based on the global features to generate coherent videos. Specifically, we propose a video auto-encoder, where a video encoder encodes videos into global features, and a video decoder, built on a diffusion model, decodes the global features and synthesizes video frames in a non-autoregressive manner. To achieve maximum flexibility, our video decoder perceives temporal information through normalized frame indexes, which enables it to synthesize arbitrary sub video clips with predetermined starting and ending frame indexes. Moreover, a novel adversarial loss is introduced to improve the global coherence and local realism between the synthesized video frames. Finally, we employ a diffusion-based video generator to fit the global features outputted by the video encoder for video generation. Extensive experimental results demonstrate the effectiveness and efficiency of our proposed method, and new state-of-the-art results have been achieved on multiple benchmarks.
Denoising diffusion probabilistic models (DDPMs) are becoming the leading paradigm for generative models. It has recently shown breakthroughs in audio synthesis, time series imputation and forecasting. In this paper, we propose Diffusion-TS, a novel diffusion-based framework that generates multivariate time series samples of high quality by using an encoder-decoder transformer with disentangled temporal representations, in which the decomposition technique guides Diffusion-TS to capture the semantic meaning of time series while transformers mine detailed sequential information from the noisy model input. Different from existing diffusion-based approaches, we train the model to directly reconstruct the sample instead of the noise in each diffusion step, combining a Fourier-based loss term. Diffusion-TS is expected to generate time series satisfying both interpretablity and realness. In addition, it is shown that the proposed Diffusion-TS can be easily extended to conditional generation tasks, such as forecasting and imputation, without any model changes. This also motivates us to further explore the performance of Diffusion-TS under irregular settings. Finally, through qualitative and quantitative experiments, results show that Diffusion-TS achieves the state-of-the-art results on various realistic analyses of time series.
Sequence-to-sequence vision-language models are showing promise, but their applicability is limited by their inference latency due to their autoregressive way of generating predictions. We propose a parallel decoding sequence-to-sequence vision-language model, trained with a Query-CTC loss, that marginalizes over multiple inference paths in the decoder. This allows us to model the joint distribution of tokens, rather than restricting to conditional distribution as in an autoregressive model. The resulting model, NARVL, achieves performance on-par with its state-of-the-art autoregressive counterpart, but is faster at inference time, reducing from the linear complexity associated with the sequential generation of tokens to a paradigm of constant time joint inference.
Robot navigation under visual corruption presents a formidable challenge. To address this, we propose a Test-time Adaptation (TTA) method, named as TTA-Nav, for point-goal navigation under visual corruptions. Our "plug-and-play" method incorporates a top-down decoder to a pre-trained navigation model. Firstly, the pre-trained navigation model gets a corrupted image and extracts features. Secondly, the top-down decoder produces the reconstruction given the high-level features extracted by the pre-trained model. Then, it feeds the reconstruction of a corrupted image back to the pre-trained model. Finally, the pre-trained model does forward pass again to output action. Despite being trained solely on clean images, the top-down decoder can reconstruct cleaner images from corrupted ones without the need for gradient-based adaptation. The pre-trained navigation model with our top-down decoder significantly enhances navigation performance across almost all visual corruptions in our benchmarks. Our method improves the success rate of point-goal navigation from the state-of-the-art result of 46% to 94% on the most severe corruption. This suggests its potential for broader application in robotic visual navigation. Project page: https://sites.google.com/view/tta-nav
In cooperative multi-agent reinforcement learning (MARL), agents aim to achieve a common goal, such as defeating enemies or scoring a goal. Existing MARL algorithms are effective but still require significant learning time and often get trapped in local optima by complex tasks, subsequently failing to discover a goal-reaching policy. To address this, we introduce Efficient episodic Memory Utilization (EMU) for MARL, with two primary objectives: (a) accelerating reinforcement learning by leveraging semantically coherent memory from an episodic buffer and (b) selectively promoting desirable transitions to prevent local convergence. To achieve (a), EMU incorporates a trainable encoder/decoder structure alongside MARL, creating coherent memory embeddings that facilitate exploratory memory recall. To achieve (b), EMU introduces a novel reward structure called episodic incentive based on the desirability of states. This reward improves the TD target in Q-learning and acts as an additional incentive for desirable transitions. We provide theoretical support for the proposed incentive and demonstrate the effectiveness of EMU compared to conventional episodic control. The proposed method is evaluated in StarCraft II and Google Research Football, and empirical results indicate further performance improvement over state-of-the-art methods.
What does a neural network learn when training from a task-specific dataset? Synthesizing this knowledge is the central idea behind Dataset Distillation, which recent work has shown can be used to compress large datasets into a small set of input-label pairs ($\textit{prototypes}$) that capture essential aspects of the original dataset. In this paper, we make the key observation that existing methods distilling into explicit prototypes are very often suboptimal, incurring in unexpected storage cost from distilled labels. In response, we propose $\textit{Distributional Dataset Distillation}$ (D3), which encodes the data using minimal sufficient per-class statistics and paired with a decoder, we distill dataset into a compact distributional representation that is more memory-efficient compared to prototype-based methods. To scale up the process of learning these representations, we propose $\textit{Federated distillation}$, which decomposes the dataset into subsets, distills them in parallel using sub-task experts and then re-aggregates them. We thoroughly evaluate our algorithm on a three-dimensional metric and show that our method achieves state-of-the-art results on TinyImageNet and ImageNet-1K. Specifically, we outperform the prior art by $6.9\%$ on ImageNet-1K under the storage budget of 2 images per class.
Feature selection is a crucial task in settings where data is high-dimensional or acquiring the full set of features is costly. Recent developments in neural network-based embedded feature selection show promising results across a wide range of applications. Concrete Autoencoders (CAEs), considered state-of-the-art in embedded feature selection, may struggle to achieve stable joint optimization, hurting their training time and generalization. In this work, we identify that this instability is correlated with the CAE learning duplicate selections. To remedy this, we propose a simple and effective improvement: Indirectly Parameterized CAEs (IP-CAEs). IP-CAEs learn an embedding and a mapping from it to the Gumbel-Softmax distributions' parameters. Despite being simple to implement, IP-CAE exhibits significant and consistent improvements over CAE in both generalization and training time across several datasets for reconstruction and classification. Unlike CAE, IP-CAE effectively leverages non-linear relationships and does not require retraining the jointly optimized decoder. Furthermore, our approach is, in principle, generalizable to Gumbel-Softmax distributions beyond feature selection.
The task-conditional model is a distinctive stream for efficient multi-task learning. Existing works encounter a critical limitation in learning task-agnostic and task-specific representations, primarily due to shortcomings in global context modeling arising from CNN-based architectures, as well as a deficiency in multi-scale feature interaction within the decoder. In this paper, we introduce a novel task-conditional framework called Task Indicating Transformer (TIT) to tackle this challenge. Our approach designs a Mix Task Adapter module within the transformer block, which incorporates a Task Indicating Matrix through matrix decomposition, thereby enhancing long-range dependency modeling and parameter-efficient feature adaptation by capturing intra- and inter-task features. Moreover, we propose a Task Gate Decoder module that harnesses a Task Indicating Vector and gating mechanism to facilitate adaptive multi-scale feature refinement guided by task embeddings. Experiments on two public multi-task dense prediction benchmarks, NYUD-v2 and PASCAL-Context, demonstrate that our approach surpasses state-of-the-art task-conditional methods.
Information extraction (IE) aims to extract complex structured information from the text. Numerous datasets have been constructed for various IE tasks, leading to time-consuming and labor-intensive data annotations. Nevertheless, most prevailing methods focus on training task-specific models, while the common knowledge among different IE tasks is not explicitly modeled. Moreover, the same phrase may have inconsistent labels in different tasks, which poses a big challenge for knowledge transfer using a unified model. In this study, we propose a regularization-based transfer learning method for IE (TIE) via an instructed graph decoder. Specifically, we first construct an instruction pool for datasets from all well-known IE tasks, and then present an instructed graph decoder, which decodes various complex structures into a graph uniformly based on corresponding instructions. In this way, the common knowledge shared with existing datasets can be learned and transferred to a new dataset with new labels. Furthermore, to alleviate the label inconsistency problem among various IE tasks, we introduce a task-specific regularization strategy, which does not update the gradients of two tasks with 'opposite direction'. We conduct extensive experiments on 12 datasets spanning four IE tasks, and the results demonstrate the great advantages of our proposed method
This paper presents the Text Encoding Diffusion Model (TEncDM), a novel approach to diffusion modeling that operates in the space of pre-trained language model encodings. In contrast to traditionally used embeddings, encodings integrate contextual information. In our approach, we also employ a transformer-based decoder, specifically designed to incorporate context in the token prediction process. We conduct a comprehensive examination of the influence of the encoder, decoder, noise scheduler, and self-conditioning on zero-shot generation. Furthermore, we compare TEncDM with previous approaches on three conditional text generation tasks: QQP, XSum, and Wiki-Auto. The results show that TEncDM exhibits superior performance compared to existing non-autoregressive diffusion models. Our code is available at https://github.com/M0RJIQUE/tencdm.