Bioacoustic recognition requires fine-grained acoustic understanding to distinguish similar-sounding species. However, many large-scale data repositories such as iNaturalist are weakly annotated, often with only a single positive species label per recording, making supervised learning particularly challenging. Inspired by advances in computer vision, recent approaches have shifted toward self-supervised learning to capture the underlying structure of audio without relying on exhaustive annotations. In particular, masked autoencoders (MAE) have shown strong transferability on massive audio corpora, yet their effectiveness in more modest bioacoustic settings remains underexplored. In this work, we conduct a systematic study of MAE pretraining for species classification on iNatSounds, analyzing the impacts of pretraining data scale, domain specificity, data curation, and transfer strategies. Consistent with prior work, we find that models pretrained on diverse general audio data achieve the best transfer performance on iNatSounds. Contrary to observations from large-scale audio benchmarks, we find that (1) additional masked reconstruction pretraining on domain-specific data provides limited benefits and may even degrade performance relative to off-the-shelf models, and (2) selective data filtering offers a negligible advantage when the overall data scale is limited. Our results indicate that, in moderate-sized fine-grained bioacoustic settings, pretraining scale dominates objective design. These findings further clarify when MAE-based pretraining is effective and provide practical guidance for model selection under limited supervision.
Masked autoencoders are scalable vision learners, as the title of MAE \cite{he2022masked}, which suggests that self-supervised learning (SSL) in vision might undertake a similar trajectory as in NLP. Specifically, generative pretext tasks with the masked prediction (e.g., BERT) have become a de facto standard SSL practice in NLP. By contrast, early attempts at generative methods in vision have been buried by their discriminative counterparts (like contrastive learning); however, the success of mask image modeling has revived the masking autoencoder (often termed denoising autoencoder in the pa…
Washington University in St. Louis（圣路易斯华盛顿大学）；University of Oxford（牛津大学）
"Masked Autoencoders (MAE) Are Scalable Vision Learners" revolutionizes the self-supervised learning method in that it not only achieves the state-of-the-art for image pre-training, but is also a milestone that bridges the gap between visual and linguistic masked autoencoding (BERT-style) pre-trainings. However, to our knowledge, to date there are no theoretical perspectives to explain the powerful expressivity of MAE. In this paper, we, for the first time, propose a unified theoretical framework that provides a mathematical understanding for MAE. Specifically, we explain the patch-based atten…
This paper studies a conceptually simple extension of Masked Autoencoders (MAE) to spatiotemporal representation learning from videos. We randomly mask out spacetime patches in videos and learn an autoencoder to reconstruct them in pixels. Interestingly, we show that our MAE method can learn strong representations with almost no inductive bias on spacetime (only except for patch and positional embeddings), and spacetime-agnostic random masking performs the best. We observe that the optimal masking ratio is as high as 90% (vs. 75% on images), supporting the hypothesis that this ratio is related…
Masked Autoencoders (MAE) based on a reconstruction task have risen to be a promising paradigm for self-supervised learning (SSL) and achieve state-of-the-art performance across different benchmark datasets. However, despite its impressive empirical success, there is still limited theoretical understanding of it. In this paper, we propose a theoretical understanding of how masking matters for MAE to learn meaningful features. We establish a close connection between MAE and contrastive learning, which shows that MAE implicit aligns the mask-induced positive pairs. Built upon this connection, we…
University of Maryland（马里兰大学）；MIT CSAIL（麻省理工学院计算机科学与人工智能实验室）；Google Research（谷歌研究院）
We introduce CAN, a simple, efficient and scalable method for self-supervised learning of visual representations. Our framework is a minimal and conceptually clean synthesis of (C) contrastive learning, (A) masked autoencoders, and (N) the noise prediction approach used in diffusion models. The learning mechanisms are complementary to one another: contrastive learning shapes the embedding space across a batch of image samples; masked autoencoders focus on reconstruction of the low-frequency spatial correlations in a single image sample; and noise prediction encourages the reconstruction of the…
This paper studies a simple extension of image-based Masked Autoencoders (MAE) to self-supervised representation learning from audio spectrograms. Following the Transformer encoder-decoder design in MAE, our Audio-MAE first encodes audio spectrogram patches with a high masking ratio, feeding only the non-masked tokens through encoder layers. The decoder then re-orders and decodes the encoded context padded with mask tokens, in order to reconstruct the input spectrogram. We find it beneficial to incorporate local window attention in the decoder, as audio spectrograms are highly correlated in lo…
Classifying bioacoustic data without individual call annotations using temporal convolutional networks and feature extractors
Sea Mammal Research Unit, School of Biology, University of St. Andrews（海洋哺乳动物研究单位，生物学系，圣安德鲁大学）；Electrical and Electronic Engineering, University of Strathclyde（电子与电气工程，斯特拉斯克莱德大学）
Self-supervised learning (SSL) enables learning useful inductive biases through utilizing pretext tasks that require no labels. The unlabeled nature of SSL makes it especially important for whole slide histopathological images (WSIs), where patch-level human annotation is difficult. Masked Autoencoders (MAE) is a recent SSL method suitable for digital pathology as it does not require negative sampling and requires little to no data augmentations. However, the domain shift between natural images and digital pathology images requires further research in designing MAE for patch-level WSIs. In thi…
Aalborg University（奥尔堡大学）；University Politehnica of Bucharest（布加勒斯特理工大学）；University of Bucharest（布加勒斯特大学）；Milestone Systems（迈尔斯通系统公司）；SecurifAI（SecurifAI公司）
本文提出 CL-MAE，利用可学习掩码模块在 MAE 自监督训练中生成由易到难的重建课程，并在 ImageNet 及五个下游任务上验证其优于 MAE 的表示学习效果。
University of Notre Dame（圣母大学）；Brandeis University（布兰迪斯大学）
Generative self-supervised learning (SSL), especially masked autoencoders, has become one of the most exciting learning paradigms and has shown great potential in handling graph data. However, real-world graphs are always heterogeneous, which poses three critical challenges that existing methods ignore: 1) how to capture complex graph structure? 2) how to incorporate various node attributes? and 3) how to encode different node positions? In light of this, we study the problem of generative SSL on heterogeneous graphs and propose HGMAE, a novel heterogeneous graph masked autoencoder model to ad…
Self-supervised learning (SSL) has been extensively explored in recent years. Particularly, generative SSL has seen emerging success in natural language processing and other AI fields, such as the wide adoption of BERT and GPT. Despite this, contrastive learning-which heavily relies on structural data augmentation and complicated training strategies-has been the dominant approach in graph SSL, while the progress of generative SSL on graphs, especially graph autoencoders (GAEs), has thus far not reached the potential as promised in other fields. In this paper, we identify and examine the issues…
Texas A&M University（德克萨斯A&M大学）；University of Georgia（佐治亚大学）；The Hong Kong Polytechnic University（香港理工大学）；Samsung Research America（三星美国研究院）；Rice University（莱斯大学）
We introduce a novel masked graph autoencoder (MGAE) framework to perform effective learning on graph structure data. Taking insights from self-supervised learning, we randomly mask a large proportion of edges and try to reconstruct these missing edges during training. MGAE has two core designs. First, we find that masking a high ratio of the input graph structure, e.g., $70\%$, yields a nontrivial and meaningful self-supervisory task that benefits downstream applications. Second, we employ a graph neural network (GNN) as an encoder to perform message propagation on the partially-masked graph.…
University of Trieste（特里埃斯特大学）；Abdus Salam International Centre for Theoretical Physics (ICTP)（阿布杜斯·萨拉姆国际理论物理学中心（ICTP））；Scuola Internazionale Superiore di Studi Avanzati (SISSA)（国际先进研究高等学院（SISSA））；University of Nova Gorica（诺瓦戈里察大学）；INFN – National Institute for Nuclear Physics（意大利国家核物理研究所（INFN））；ICSC - Centro Nazionale di Ricerca in High Performance Computing（高性能计算国家研究中心（ICSC））；Imperial College London（伦敦帝国理工学院）
Wearable sensors provide abundant physiological time series, yet the principles governing their predictive utility remain unclear. We hypothesize that temporal resolution is a fundamental axis of representation learning, with different clinical and behavioral outcomes relying on structure at distinct scales. To test this resolution hypothesis, we introduce HiMAE (Hierarchical Masked Autoencoder), a self supervised framework that combines masked autoencoding with a hierarchical convolutional encoder decoder. HiMAE produces multi resolution embeddings that enable systematic evaluation of which t…
MIT CSAIL（麻省理工学院计算机科学与人工智能实验室）；UT Austin（德克萨斯大学奥斯汀分校）；IBM Research AI（IBM研究院人工智能部门）；MIT-IBM Watson AI Lab（麻省理工学院-IBM沃森人工智能实验室）；Goethe University Frankfurt（法兰克福大学）
NTT Communication Science Laboratories（NTT通信科学实验室）；NTT Corporation（NTT公司）
Recent general-purpose audio representations show state-of-the-art performance on various audio tasks. These representations are pre-trained by self-supervised learning methods that create training signals from the input. For example, typical audio contrastive learning uses temporal relationships among input sounds to create training signals, whereas some methods use a difference among input views created by data augmentations. However, these training signals do not provide information derived from the intact input sound, which we think is suboptimal for learning representation that describes …
