In this paper, we conduct a comprehensive analysis of two dual-branch (Siamese architecture) self-supervised learning approaches, namely Barlow Twins and spectral contrastive learning, through the lens of matrix mutual information. We prove that the loss functions of these methods implicitly optimize both matrix mutual information and matrix joint entropy. This insight prompts us to further explore the category of single-branch algorithms, specifically MAE and U-MAE, for which mutual information and joint entropy become the entropy. Building on this intuition, we introduce the Matrix Variational Masked Auto-Encoder (M-MAE), a novel method that leverages the matrix-based estimation of entropy as a regularizer and subsumes U-MAE as a special case. The empirical evaluations underscore the effectiveness of M-MAE compared with the state-of-the-art methods, including a 3.9% improvement in linear probing ViT-Base, and a 1% improvement in fine-tuning ViT-Large, both on ImageNet.
Understanding point clouds captured from the real-world is challenging due to shifts in data distribution caused by varying object scales, sensor angles, and self-occlusion. Prior works have addressed this issue by combining recent learning principles such as self-supervised learning, self-training, and adversarial training, which leads to significant computational overhead.Toward succinct yet powerful domain adaptation for point clouds, we revisit the unique challenges of point cloud data under domain shift scenarios and discover the importance of the global geometry of source data and trends of target pseudo-labels biased to the source label distribution. Motivated by our observations, we propose an adapter-guided domain adaptation method, PC-Adapter, that preserves the global shape information of the source domain using an attention-based adapter, while learning the local characteristics of the target domain via another adapter equipped with graph convolution. Additionally, we propose a novel pseudo-labeling strategy resilient to the classifier bias by adjusting confidence scores using their class-wise confidence distributions to consider relative confidences. Our method demonstrates superiority over baselines on various domain shift settings in benchmark datasets - PointDA, GraspNetPC, and PointSegDA.
Recommendation systems rely on historical clicks to learn user interests and provide appropriate items. However, current studies tend to treat clicks equally, which may ignore the assorted intensities of user interests in different clicks. In this paper, we aim to achieve multi-granularity Click confidence Learning via Self-Distillation in recommendation (CLSD). Due to the lack of supervised signals in click confidence, we first apply self-supervised learning to obtain click confidence scores via a global self-distillation method. After that, we define a local confidence function to adapt confidence scores at the user group level, since the confidence distributions can be varied among user groups. With the combination of multi-granularity confidence learning, we can distinguish the quality of clicks and model user interests more accurately without involving extra data and model structures. The significant improvements over different backbones on industrial offline and online experiments in a real-world recommender system prove the effectiveness of our model. Recently, CLSD has been deployed on a large-scale recommender system, affecting over 400 million users.
Speech signals, typically sampled at rates in the tens of thousands per second, contain redundancies, evoking inefficiencies in sequence modeling. High-dimensional speech features such as spectrograms are often used as the input for the subsequent model. However, they can still be redundant. Recent investigations proposed the use of discrete speech units derived from self-supervised learning representations, which significantly compresses the size of speech data. Applying various methods, such as de-duplication and subword modeling, can further compress the speech sequence length. Hence, training time is significantly reduced while retaining notable performance. In this study, we undertake a comprehensive and systematic exploration into the application of discrete units within end-to-end speech processing models. Experiments on 12 automatic speech recognition, 3 speech translation, and 1 spoken language understanding corpora demonstrate that discrete units achieve reasonably good results in almost all the settings. We intend to release our configurations and trained models to foster future research efforts.
Unsupervised domain adaptation aims to transfer knowledge from a fully-labeled source domain to an unlabeled target domain. However, in real-world scenarios, providing abundant labeled data even in the source domain can be infeasible due to the difficulty and high expense of annotation. To address this issue, recent works consider the Few-shot Unsupervised Domain Adaptation (FUDA) where only a few source samples are labeled, and conduct knowledge transfer via self-supervised learning methods. Yet existing methods generally overlook that the sparse label setting hinders learning reliable source knowledge for transfer. Additionally, the learning difficulty difference in target samples is different but ignored, leaving hard target samples poorly classified. To tackle both deficiencies, in this paper, we propose a novel Confidence-based Visual Dispersal Transfer learning method (C-VisDiT) for FUDA. Specifically, C-VisDiT consists of a cross-domain visual dispersal strategy that transfers only high-confidence source knowledge for model adaptation and an intra-domain visual dispersal strategy that guides the learning of hard target samples with easy ones. We conduct extensive experiments on Office-31, Office-Home, VisDA-C, and DomainNet benchmark datasets and the results demonstrate that the proposed C-VisDiT significantly outperforms state-of-the-art FUDA methods. Our code is available at https://github.com/Bostoncake/C-VisDiT.
This study proposes a pipeline that incorporates a novel style transfer model and a simultaneous super-resolution and segmentation model. The proposed pipeline aims to enhance diffusion tensor imaging (DTI) images by translating them into the late gadolinium enhancement (LGE) domain, which offers a larger amount of data with high-resolution and distinct highlighting of myocardium infarction (MI) areas. Subsequently, the segmentation task is performed on the LGE style image. An end-to-end super-resolution segmentation model is introduced to generate high-resolution mask from low-resolution LGE style DTI image. Further, to enhance the performance of the model, a multi-task self-supervised learning strategy is employed to pre-train the super-resolution segmentation model, allowing it to acquire more representative knowledge and improve its segmentation performance after fine-tuning. https: github.com/wlc2424762917/Med_Img
Human anatomy is the foundation of medical imaging and boasts one striking characteristic: its hierarchy in nature, exhibiting two intrinsic properties: (1) locality: each anatomical structure is morphologically distinct from the others; and (2) compositionality: each anatomical structure is an integrated part of a larger whole. We envision a foundation model for medical imaging that is consciously and purposefully developed upon this foundation to gain the capability of "understanding" human anatomy and to possess the fundamental properties of medical imaging. As our first step in realizing this vision towards foundation models in medical imaging, we devise a novel self-supervised learning (SSL) strategy that exploits the hierarchical nature of human anatomy. Our extensive experiments demonstrate that the SSL pretrained model, derived from our training strategy, not only outperforms state-of-the-art (SOTA) fully/self-supervised baselines but also enhances annotation efficiency, offering potential few-shot segmentation capabilities with performance improvements ranging from 9% to 30% for segmentation tasks compared to SSL baselines. This performance is attributed to the significance of anatomy comprehension via our learning strategy, which encapsulates the intrinsic attributes of anatomical structures-locality and compositionality-within the embedding space, yet overlooked in existing SSL methods. All code and pretrained models are available at https://github.com/JLiangLab/Eden.
Motion prediction is crucial for autonomous vehicles to operate safely in complex traffic environments. Extracting effective spatiotemporal relationships among traffic elements is key to accurate forecasting. Inspired by the successful practice of pretrained large language models, this paper presents SEPT, a modeling framework that leverages self-supervised learning to develop powerful spatiotemporal understanding for complex traffic scenes. Specifically, our approach involves three masking-reconstruction modeling tasks on scene inputs including agents' trajectories and road network, pretraining the scene encoder to capture kinematics within trajectory, spatial structure of road network, and interactions among roads and agents. The pretrained encoder is then finetuned on the downstream forecasting task. Extensive experiments demonstrate that SEPT, without elaborate architectural design or manual feature engineering, achieves state-of-the-art performance on the Argoverse 1 and Argoverse 2 motion forecasting benchmarks, outperforming previous methods on all main metrics by a large margin.
We present a novel objective function for cluster-based self-supervised learning (SSL) that is designed to circumvent the triad of failure modes, namely representation collapse, cluster collapse, and the problem of invariance to permutations of cluster assignments. This objective consists of three key components: (i) A generative term that penalizes representation collapse, (ii) a term that promotes invariance to data augmentations, thereby addressing the issue of label permutations and (ii) a uniformity term that penalizes cluster collapse. Additionally, our proposed objective possesses two notable advantages. Firstly, it can be interpreted from a Bayesian perspective as a lower bound on the data log-likelihood. Secondly, it enables the training of a standard backbone architecture without the need for asymmetric elements like stop gradients, momentum encoders, or specialized clustering layers. Due to its simplicity and theoretical foundation, our proposed objective is well-suited for optimization. Experiments on both toy and real world data demonstrate its effectiveness
Sparse keypoint matching based on distinct 3D feature representations can improve the efficiency and robustness of point cloud registration. Existing learning-based 3D descriptors and keypoint detectors are either independent or loosely coupled, so they cannot fully adapt to each other. In this work, we propose a tightly coupled keypoint detector and descriptor (TCKDD) based on a multi-task fully convolutional network with a probabilistic detection loss. In particular, this self-supervised detection loss fully adapts the keypoint detector to any jointly learned descriptors and benefits the self-supervised learning of descriptors. Extensive experiments on both indoor and outdoor datasets show that our TCKDD achieves state-of-the-art performance in point cloud registration. Furthermore, we design a keypoint detector and descriptors-assisted LiDAR odometry and mapping framework (KDD-LOAM), whose real-time odometry relies on keypoint descriptor matching-based RANSAC. The sparse keypoints are further used for efficient scan-to-map registration and mapping. Experiments on KITTI dataset demonstrate that KDD-LOAM significantly surpasses LOAM and shows competitive performance in odometry.
Multilingual self-supervised learning (SSL) has often lagged behind state-of-the-art (SOTA) methods due to the expenses and complexity required to handle many languages. This further harms the reproducibility of SSL, which is already limited to few research groups due to its resource usage. We show that more powerful techniques can actually lead to more efficient pre-training, opening SSL to more research groups. We propose WavLabLM, which extends WavLM's joint prediction and denoising to 40k hours of data across 136 languages. To build WavLabLM, we devise a novel multi-stage pre-training method, designed to address the language imbalance of multilingual data. WavLabLM achieves comparable performance to XLS-R on ML-SUPERB with less than 10% of the training data, making SSL realizable with academic compute. We show that further efficiency can be achieved with a vanilla HuBERT Base model, which can maintain 94% of XLS-R's performance with only 3% of the data, 4 GPUs, and limited trials. We open-source all code and models in ESPnet.
We present a new pre-training strategy called M$^{3}$3D ($\underline{M}$ulti-$\underline{M}$odal $\underline{M}$asked $\underline{3D}$) built based on Multi-modal masked autoencoders that can leverage 3D priors and learned cross-modal representations in RGB-D data. We integrate two major self-supervised learning frameworks; Masked Image Modeling (MIM) and contrastive learning; aiming to effectively embed masked 3D priors and modality complementary features to enhance the correspondence between modalities. In contrast to recent approaches which are either focusing on specific downstream tasks or require multi-view correspondence, we show that our pre-training strategy is ubiquitous, enabling improved representation learning that can transfer into improved performance on various downstream tasks such as video action recognition, video action detection, 2D semantic segmentation and depth estimation. Experiments show that M$^{3}$3D outperforms the existing state-of-the-art approaches on ScanNet, NYUv2, UCF-101 and OR-AR, particularly with an improvement of +1.3\% mIoU against Mask3D on ScanNet semantic segmentation. We further evaluate our method on low-data regime and demonstrate its superior data efficiency compared to current state-of-the-art approaches.
Recent advances with self-supervised learning have allowed speech recognition systems to achieve state-of-the-art (SOTA) word error rates (WER) while requiring only a fraction of the labeled training data needed by its predecessors. Notwithstanding, while such models achieve SOTA performance in matched train/test conditions, their performance degrades substantially when tested in unseen conditions. To overcome this problem, strategies such as data augmentation and/or domain shift training have been explored. Available models, however, are still too large to be considered for edge speech applications on resource-constrained devices, thus model compression tools are needed. In this paper, we explore the effects that train/test mismatch conditions have on speech recognition accuracy based on compressed self-supervised speech models. In particular, we report on the effects that parameter quantization and model pruning have on speech recognition accuracy based on the so-called robust wav2vec 2.0 model under noisy, reverberant, and noise-plus-reverberation conditions.
We develop techniques for refining representations for fine-grained classification and segmentation tasks in a self-supervised manner. We find that fine-tuning methods based on instance-discriminative contrastive learning are not as effective, and posit that recognizing part-specific variations is crucial for fine-grained categorization. We present an iterative learning approach that incorporates part-centric equivariance and invariance objectives. First, pixel representations are clustered to discover parts. We analyze the representations from convolutional and vision transformer networks that are best suited for this task. Then, a part-centric learning step aggregates and contrasts representations of parts within an image. We show that this improves the performance on image classification and part segmentation tasks across datasets. For example, under a linear-evaluation scheme, the classification accuracy of a ResNet50 trained on ImageNet using DetCon, a self-supervised learning approach, improves from 35.4% to 42.0% on the Caltech-UCSD Birds, from 35.5% to 44.1% on the FGVC Aircraft, and from 29.7% to 37.4% on the Stanford Cars. We also observe significant gains in few-shot part segmentation tasks using the proposed technique, while instance-discriminative learning was not as effective. Smaller, yet consistent, improvements are also observed for stronger networks based on transformers.
In the field of medical image analysis, achieving high accuracy is not enough; ensuring well-calibrated predictions is also crucial. Confidence scores of a deep neural network play a pivotal role in explainability by providing insights into the model's certainty, identifying cases that require attention, and establishing trust in its predictions. Consequently, the significance of a well-calibrated model becomes paramount in the medical imaging domain, where accurate and reliable predictions are of utmost importance. While there has been a significant effort towards training modern deep neural networks to achieve high accuracy on medical imaging tasks, model calibration and factors that affect it remain under-explored. To address this, we conducted a comprehensive empirical study that explores model performance and calibration under different training regimes. We considered fully supervised training, which is the prevailing approach in the community, as well as rotation-based self-supervised method with and without transfer learning, across various datasets and architecture sizes. Multiple calibration metrics were employed to gain a holistic understanding of model calibration. Our study reveals that factors such as weight distributions and the similarity of learned representations correlate with the calibration trends observed in the models. Notably, models trained using rotation-based self-supervised pretrained regime exhibit significantly better calibration while achieving comparable or even superior performance compared to fully supervised models across different medical imaging datasets. These findings shed light on the importance of model calibration in medical image analysis and highlight the benefits of incorporating self-supervised learning approach to improve both performance and calibration.
This paper explores Large Batch Training techniques using layer-wise adaptive scaling ratio (LARS) across diverse settings, uncovering insights. LARS algorithms with warm-up tend to be trapped in sharp minimizers early on due to redundant ratio scaling. Additionally, a fixed steep decline in the latter phase restricts deep neural networks from effectively navigating early-phase sharp minimizers. Building on these findings, we propose Time Varying LARS (TVLARS), a novel algorithm that replaces warm-up with a configurable sigmoid-like function for robust training in the initial phase. TVLARS promotes gradient exploration early on, surpassing sharp optimizers and gradually transitioning to LARS for robustness in later phases. Extensive experiments demonstrate that TVLARS consistently outperforms LARS and LAMB in most cases, with up to 2\% improvement in classification scenarios. Notably, in all self-supervised learning cases, TVLARS dominates LARS and LAMB with performance improvements of up to 10\%.
Recent years have witnessed significant advancements in self-supervised learning (SSL) methods for speech-processing tasks. Various speech-based SSL models have been developed and present promising performance on a range of downstream tasks including speech recognition. However, existing speech-based SSL models face a common dilemma in terms of computational cost, which might hinder their potential application and in-depth academic research. To address this issue, we first analyze the computational cost of different modules during HuBERT pre-training and then introduce a stack of efficiency optimizations, which is named Fast-HuBERT in this paper. The proposed Fast-HuBERT can be trained in 1.1 days with 8 V100 GPUs on the Librispeech 960h benchmark, without performance degradation, resulting in a 5.2x speedup, compared to the original implementation. Moreover, we explore two well-studied techniques in the Fast-HuBERT and demonstrate consistent improvements as reported in previous work.
This research introduces an enhanced version of the multi-objective speech assessment model--MOSA-Net+, by leveraging the acoustic features from Whisper, a large-scaled weakly supervised model. We first investigate the effectiveness of Whisper in deploying a more robust speech assessment model. After that, we explore combining representations from Whisper and SSL models. The experimental results reveal that Whisper's embedding features can contribute to more accurate prediction performance. Moreover, combining the embedding features from Whisper and SSL models only leads to marginal improvement. As compared to intrusive methods, MOSA-Net, and other SSL-based speech assessment models, MOSA-Net+ yields notable improvements in estimating subjective quality and intelligibility scores across all evaluation metrics in Taiwan Mandarin Hearing In Noise test - Quality & Intelligibility (TMHINT-QI) dataset. To further validate its robustness, MOSA-Net+ was tested in the noisy-and-enhanced track of the VoiceMOS Challenge 2023, where it obtained the top-ranked performance among nine systems.
While image data starts to enjoy the simple-but-effective self-supervised learning scheme built upon masking and self-reconstruction objective thanks to the introduction of tokenization procedure and vision transformer backbone, convolutional neural networks as another important and widely-adopted architecture for image data, though having contrastive-learning techniques to drive the self-supervised learning, still face the difficulty of leveraging such straightforward and general masking operation to benefit their learning process significantly. In this work, we aim to alleviate the burden of including masking operation into the contrastive-learning framework for convolutional neural networks as an extra augmentation method. In addition to the additive but unwanted edges (between masked and unmasked regions) as well as other adverse effects caused by the masking operations for ConvNets, which have been discussed by prior works, we particularly identify the potential problem where for one view in a contrastive sample-pair the randomly-sampled masking regions could be overly concentrated on important/salient objects thus resulting in misleading contrastiveness to the other view. To this end, we propose to explicitly take the saliency constraint into consideration in which the masked regions are more evenly distributed among the foreground and background for realizing the masking-based augmentation. Moreover, we introduce hard negative samples by masking larger regions of salient patches in an input image. Extensive experiments conducted on various datasets, contrastive learning mechanisms, and downstream tasks well verify the efficacy as well as the superior performance of our proposed method with respect to several state-of-the-art baselines.
To integrate action recognition into autonomous robotic systems, it is essential to address challenges such as person occlusions-a common yet often overlooked scenario in existing self-supervised skeleton-based action recognition methods. In this work, we propose IosPSTL, a simple and effective self-supervised learning framework designed to handle occlusions. IosPSTL combines a cluster-agnostic KNN imputer with an Occluded Partial Spatio-Temporal Learning (OPSTL) strategy. First, we pre-train the model on occluded skeleton sequences. Then, we introduce a cluster-agnostic KNN imputer that performs semantic grouping using k-means clustering on sequence embeddings. It imputes missing skeleton data by applying K-Nearest Neighbors in the latent space, leveraging nearby sample representations to restore occluded joints. This imputation generates more complete skeleton sequences, which significantly benefits downstream self-supervised models. To further enhance learning, the OPSTL module incorporates Adaptive Spatial Masking (ASM) to make better use of intact, high-quality skeleton sequences during training. Our method achieves state-of-the-art performance on the occluded versions of the NTU-60 and NTU-120 datasets, demonstrating its robustness and effectiveness under challenging conditions. Code is available at https://github.com/cyfml/OPSTL.